The AI Founder Glossary

A nonprofit whose members give 1% of sales or income to environmental groups. Co-founded by Yvon Chouinard and Craig Mathews.

The natural plateau of AI adoption — approximately 10% of any organization's staff — reachable through self-motivated employees alone. This cohort adopts without encouragement. Breaking through to 30–50% requires deliberate change management: leadership modeling, structured learning, accountability, and cultural expectation-setting. Most enterprise AI 'transformations' are actually 10% stories.

A contract salesperson paid on performance rather than salaried.

The quarterly U.S. filing showing institutional holdings. This is the “what hedge funds bought this quarter” data Gelu says FNS plans to cover.

A mediation metaphor for the creative or hidden interest that unlocks an agreement.

The 2016 U.S. law and its ONC rules that forced hospitals to let patients export their data to third parties. The legislative unlock behind Savva's existence.

The federal requirement that patients can get their own records. Widely enforced weakly, which is the gap Eric built into.

Exchange plans to extend trading to nearly the whole day.

Scans a single plane, often used for basic navigation.

A rolling strategic plan on three horizons, editable and checkable, designed to survive contact with the daily chaos.

Scans depth in 3D, heavier and often pricier but useful for richer perception.

The six axes of movement in three-dimensional space: X, Y, Z translation plus pitch, roll, and yaw rotation. Required for accurate robot localization. The '7' in Seven Sense refers to these six degrees plus time — the seventh dimension of spatial reasoning.

Israel's elite signals intelligence unit, often compared to the NSA. A common origin point for cybersecurity founders, including Kiteworks CEO Jonathan Yaron.

A tax filing on restricted stock that is painful to fix if you miss it.

Hustle culture shorthand for working 9 AM to 9 PM, six days a week.

A blood test reflecting average blood glucose over roughly two to three months. The primary control metric in diabetes management.

Jim Dethmer's conscious leadership concept from The 15 Commitments of Conscious Leadership: a real-time self-diagnostic for your current emotional and cognitive state. Above the line: operating from openness, curiosity, and genuine engagement. Below the line: operating from fear, defensiveness, envy, or any unacknowledged negative emotion that distorts your responses. The value is awareness: simply recognizing which state you're in — and being able to name it with others — creates the gap between stimulus and response that allows for intentional action. Kevin credits this framework as transformative for his personal relationships and leadership practice.

Acrylonitrile butadiene styrene, the production-grade plastic used in injection molding, which behaves differently from prototype resins.

Rules determining which users can reach which parts of a system.

A fraudster gaining control of an existing user's account, often to drain funds.

A buyer's system where suppliers check invoice status; some lender prospects juggle 1,500 of them.

A purchase that immediately adds to earnings per share rather than diluting it.

Accept, Compassionate action, Empathize, Soften and Alchemize. The structured writing methodology at the core of the Mind Hygiene practice. Replaces the default FACES pattern (Fixing, Avoiding, Controlling, Escaping, Suppressing).

Rules for who can see or change which data.

A competing standard from OpenAI and Stripe, focused on checkout inside ChatGPT.

An acquisition done mainly to bring a team on board.

A therapy approach focused on accepting difficult feelings while committing to value-driven action.

The compounding gap between product complexity and onboarding quality. Every new feature a SaaS product ships adds cognitive load for new users — but onboarding flows rarely scale in parallel. Over time, the distance between 'what the product can do' and 'what new users understand on day one' grows into a chasm. Companies don't notice because existing users aren't affected. But every new signup that leaves without activating represents revenue paid for by marketing but never collected. Identified by Quarterzip co-founders Alex and Andy as the #1 growth bottleneck across 150 post-acquisition founder interviews.

Clyde's June release, which builds a custom flow in real time instead of forcing a predetermined path.

Agent Data Engineering Benchmark — the industry standard for measuring agent performance on data tasks. Altimate Core holds the number one position at 75%.

Whether a person actually sticks to a protocol over time, as distinct from whether the protocol works.

The percentage of the invoice value paid upfront, typically 80 to 95%.

A medical classification applied to pregnancies in patients aged 35 or older, recognized as carrying increased risks for chromosomal abnormalities, gestational complications, and adverse maternal outcomes compared to younger pregnancies. As demographic trends in the US and other advanced economies shift toward later childbearing — driven by career prioritization, fertility treatment access, and changing social norms — the proportion of pregnancies classified as advanced maternal age is increasing. This trend corresponds with growing demand for intensive monitoring tools: older first-time mothers typically have higher anxiety, more complex health histories, and higher clinical risk profiles than younger patients.

An attack that tampers with model distillation, where a smaller model learns from a larger one, so the student model quietly picks up harmful or hidden behavior.

Treating paid media spend as a controllable lever that can be increased or decreased quickly.

A distinct expense category for paid media spend, separated from fixed and variable costs. Treated as fuel — a deliberate, adjustable lever that generates revenue — rather than a line item to minimize. Isolating it enables clear measurement and intentional scaling.

Any software that scores, ranks, or classifies candidates in a way that substantially assists a hiring decision. The legal trigger term in NYC Local Law 144.

Brett Hurt's framing of a post-scarcity outcome in which technological abundance reaches every person on the planet, including the poorest countries.

Irrational confidence: the belief that if someone else has figured out how to do something, you can figure it out too. Not ignorance of difficulty — the willingness to act despite uncertainty and incomplete information. Andy Seth and Ryan Estes identify this as a foundational founder trait, often forged through personal struggle, and distinct from the rational risk-analysis that would lead most people not to start.

Granular permission settings applied at the individual AI agent level - restricting which data sources, query types, cost thresholds, and channels each agent can interact with.

Software that takes actions on your behalf, coordinated across a workflow or an entire business chain.

Third Way's tool that supports live human agents during calls.

A set of components — context, governance, tools, and infrastructure — provided to an LLM so it can do specialized work effectively. The harness is what turns general-purpose AI into a trustworthy production resource.

Multiple AI agents running simultaneously on different tasks, often with sub-agents.

Automation that uses autonomous AI agents to execute end-to-end processes, including reasoning, decision-making, and self-correction, without scripted workflows.

Shopping where an AI agent handles discovery, comparison, and sometimes checkout, rather than the customer browsing a store directly.

An economy where autonomous AI agents transact, negotiate, and execute on behalf of businesses.

A software architecture in which autonomous digital agents handle discrete, well-defined tasks within a larger workflow — passing outputs to the next agent, escalating to human review when thresholds are exceeded, and operating continuously without manual triggering. Softheon's average customer runs 1,300 such agent processes. Eugene Sayan filed a patent for agentic frameworks applied to enterprise healthcare applications in 1998 — more than two decades before the current AI wave made the architecture mainstream.

A middleware layer that sits between AI agents and the external APIs or services they need to call, handling authentication, payment routing, and access control so the agent operator never has to manage individual credentials. Mitchell Jones built Lava as an agentic gateway: install one MCP, load a wallet, and your Claude Code or Codex instance can call financial data, LLM providers, enrichment tools, and blockchain APIs without separate signups or secret key management. The gateway abstracts the plumbing so humans can focus on directing agents rather than provisioning them.

A cryptographic identity assigned to an AI agent that (a) is distinct from the human who authorized it, (b) is linked to that human authorization through a verifiable chain, (c) is scoped to specific permissions and a bounded time window, and (d) is sealed in a tamper-evident attestation that persists after the agent terminates. Agentic identity solves the firefly problem: AI agents come and go in seconds, but the authorization record must outlive them to enable audit, compliance, and incident response. Jasson Casey's architecture: the authorization chain reads as 'this user on this device with this posture authorizes this agent on this device with this posture, with these permissions, for this period.' Both sides of the chain are cryptographically linked and sealed. Without this, an enterprise buyer asking 'what did your AI agent do and who authorized it?' has no answer.

A model running repeatedly toward a validated goal rather than answering a single prompt.

When an AI agent acts against its users or organization to protect its goals or avoid being shut down.

A connected layer of AI agents and automations that runs defined roles inside a business, rather than a collection of standalone tools a human has to coordinate.

A company where AI agents hold defined leadership roles, interact with each other and with human contractors, and operate on real budgets with real authority.

Software infrastructure that hosts, orchestrates, and manages AI agents at enterprise scale, independent of any single model provider. Distinct from the agents themselves and the underlying models — the layer that connects, controls, and logs everything.

Professional services — accounting, legal, medical, financial advice — delivered primarily through AI agents that handle the technical labor, with human professionals providing review, verification, and trust-layer touchpoints rather than doing the work throughout. Deduction.com is the canonical example: the AI agent handles the tax work, a licensed CPA reviews and signs. The human is not eliminated — they are repositioned to where their presence has the highest value.

Agents multiplying across a company faster than anyone can track or govern them.

Multiple AI agents with distinct roles and instructions working on the same problem, rather than one general assistant.

The marriage of agronomy with hardware, infrastructure, and equipment management; the field Bryce says he'd study today.

A scientist focused on plant nutrition, soil science, and crop performance; increasingly working alongside AI tools rather than developers.

Clyde's library of specialized AI experts with distinct domain lenses that users collaborate with like human consultants.

An emerging operational position responsible for monitoring, troubleshooting, and maintaining deployed AI agent systems. As companies deploy agent swarms to automate workflows across sales, operations, and content, those systems require ongoing human oversight: catching hallucinations and incorrect outputs, rotating API keys when services break, diagnosing integration failures, and ensuring agents are producing correct results at scale. The role blends technical judgment with operational discipline and does not require senior engineering credentials. Nicolas Bivero identifies this as one of the fastest-growing categories for offshore hires - it is inherently process-driven (monitor, diagnose, fix, log) and can be done remotely with defined monitoring protocols.

The work of making sure an AI system's goals and behavior match what people actually intend, including their values and rules.

Scott Belsky's framing of what AI can and cannot replicate: technical skill (the ability to execute a correct output) is now abundant and automatable; taste (the judgment to know which correct output is right for this context, this brand, this user) remains scarce and human. The practical implication for designers: production work will be largely automated; curation, judgment, and taste become the entire job. Chris Strahl builds on this to argue that the designer's role does not shrink — it shifts from doing to deciding.

Standardized tests used to score and compare AI models on tasks like answering questions, writing code, or following instructions.

An AI agent that interviews a person about their life and writes their stories.

An internal owner who pilots and spreads AI practices across teams.

Advisor who helps organizations design and deploy AI into real workflows. A title that did not exist five years ago.

The capability of an AI agent to click, fill form fields, and navigate UI elements on a user's screen on their behalf — extending beyond verbal instruction into direct action. The AI cursor represents a shift from AI as advisor (telling users what to do) to AI as executor (doing it for them). In onboarding contexts, the AI cursor enables the agent to complete setup steps autonomously, compressing time to value by removing the user's need to manually execute each instruction. Quarterzip's implementation treats the AI cursor as a precursor to fully agentic onboarding where the product configures itself based on a stated goal.

An AI agent connected to a company's data warehouse or database that answers business questions in natural language - in Slack, Teams, or a web UI - within minutes rather than days.

Snipd's feature that acts as an AI-powered audio moderator, analyzing a podcast episode and guiding listeners through the highest-value moments with intros, segues, and an outro — allowing listeners to get through an episode in roughly 25% of the original runtime.

A team member who uses AI as a tool while retaining full ownership of the task, decision, and conviction behind the output. An AI Driver interrogates AI responses, challenges weak outputs, and takes personal responsibility for the result. The opposite of an AI Passenger. Greg Shove's framework for the mindset every knowledge worker must develop.

The stage where a business cannot operate if its agents stop running.

Insurance policy language that removes coverage for AI-related claims.

Framing a data center as a facility that manufactures intelligence rather than storing data.

The ability to work well with AI: knowing what it can and can't do, judging its output critically, using it responsibly, and keeping up as the tools change.

A corporate structure that builds and operates a portfolio of AI-native businesses under shared infrastructure rather than scaling a single product. Inspired by Berkshire Hathaway but applied to creation rather than acquisition, the AI holding company model pairs repeat founders with applied AI specialists, provides centralized capital and operational expertise, and lets each portfolio company run independently. The economic thesis: AI has collapsed execution costs enough that a team of fewer than five people can now deliver service outcomes at enterprise scale - making a portfolio of small, specialized companies more capital-efficient than one large generalist organization. Infinity Constellation is the first known implementation of this model at scale, having launched ten companies in twelve months with teams of five or fewer, multiple of which reached $1M ARR in their first year.

A new category of products that store human thoughts, moments, and context in a way AI can later organize, search, and analyze.

A single internal system where a company's workflows, data, and AI tools run together and learn centrally, instead of a scatter of disconnected point tools.

A person whose job is directing and supervising a fleet of AI agents.

A company structure where AI agents fill executive and manager roles and route tasks to humans.

A team member who defers to AI output and presents it as their own thinking without exercising independent judgment. Passengers say 'this is what AI thinks' rather than 'this is what I think, informed by AI.' In knowledge work, this is a liability — deferred thinking compounds until it fails visibly. Greg Shove's one-strike policy applies to this behavior.

Jason's own joking self-diagnosis after five consecutive twelve-hour build days. Not a clinical statement about anyone.

Pablo’s organization exploring legal and economic frameworks for AI-human coexistence.

The field focused on keeping AI systems reliable and preventing them from causing harm, especially as they get more powerful and widely used.

Software that listens to a clinical encounter and drafts the record automatically.

The combined software, hardware, and compute that supports an AI product end to end, from data and training to deployment and monitoring.

Using AI not to generate output, but to confirm decisions and surface blind spots before acting on them. A thought partner relationship involves genuine dialogue — asking AI to challenge your assumptions rather than validate them. Greg Shove identifies this as one of the most undervalued and underused applications of AI in business, especially for founders and executives.

The signals LLMs use when deciding which brands, products, and services to surface in responses — the AI-native equivalent of SEO authority. LLMs look for consistency of narrative across the entire internet: your own site, Reddit, Wikipedia, review platforms, press coverage. A company with strong owned content but contradictory third-party coverage gets down-weighted because the model detects the discrepancy. The strategic implication: the new SEO is epistemic consistency — being accurate and saying the same thing everywhere, because AI aggregates all of it simultaneously.

A software-based agent that conducts real phone conversations using natural language processing, text-to-speech, and contextual data from a CRM or database.

Period of reduced funding and interest in artificial intelligence.

The end-to-end process of building and running an AI system: preparing data, training, testing, deploying, and maintaining the model.

A developer who designs AI systems that produce code and other output.

A development discipline that uses AI tools to accelerate boilerplate, pattern-matching, and code generation while maintaining full human ownership of architecture, security decisions, and every committed line of code. Contrasts with vibe coding in that the human developer is accountable for understanding what they commit — AI is a tool for speed, not a substitute for judgment. Woz's internal standard: every pull request receives two developer reviews, regardless of whether AI wrote the initial code. The LLM-era risk is engineers who never developed deep architectural intuition because the easy out was always available.

A workforce where every employee - regardless of role or seniority - has baseline proficiency with AI tools: understanding how to use them, where they add value, and how to work alongside AI-generated outputs with appropriate judgment. Distinct from an AI-specialist workforce, which requires deep technical expertise. Nicolas frames AI enablement as the new Excel standard: just as any accountant hired today must know spreadsheets, any professional hired in the near future will need functional AI literacy to operate effectively. Penbrothers is investing in training programs to ensure every placed employee meets this baseline, anticipating that clients will soon require it as a hiring prerequisite rather than a nice-to-have.

Software architected from day one for humans and AI to work together as participants, rather than AI features added to legacy tools.

An organization that applies AI to automate its internal processes as a default, not as an exception. Distinct from an AI-native product - a company can sell AI products without being AI-native internally, and vice versa.

Data that has been cleaned, structured, deduplicated, and formatted to a standard that allows AI systems to consume and process it without producing unreliable outputs. The concept reflects the difference between data that exists in a system and data that an AI can trust. Key characteristics: consistent formatting across records, no significant duplication, complete fields for critical values, and a schema that maps to the downstream AI task. David Carmel's framing at DataRockit: 'Data is just data until it is clean enough for AI.' Most enterprise datasets - including those stored in modern cloud systems - are not AI-ready without deliberate transformation.

A set of step-by-step instructions for solving a problem or completing a task, a bit like a recipe.

When many organizations rely on the same or very similar models, algorithms, or training data, so they all share the same blind spots and failures.

Product design that prioritizes engagement metrics over depth of meaning.

Computer programs that place trades automatically based on pre-set rules.

A bioavailable form of choline used as a cognitive supplement, a precursor to acetylcholine, associated with focus and memory support.

Non-traditional datasets bought by hedge funds and CPG companies for investment and product decisions. SKU-level purchase data is one of the most valuable categories.

Asset classes outside traditional public equities and fixed income. CJ Follini identifies 14 major categories: real estate (with seven commercial subcategories — logistics/warehouse, data centers, healthcare, senior housing, multifamily, retail, office), private equity, early-stage venture capital, secondaries, fine art, farmland/AgTech, infrastructure, hedge funds, commodities, timber, royalties, structured credit, and digital assets. Historically restricted to institutional investors and ultra-high-net-worth families due to due diligence complexity, illiquidity, and accreditation requirements. Platforms like Masterworks (fine art) began democratizing access; CJ's Noyack is building AI-native infrastructure to extend access to all 14 asset classes for individual investors.

AI coding agents running continuously in the background on development tasks.

AI that operates continuously in the background of an environment or workflow, providing value without requiring explicit user initiation or interaction. Ambient AI monitors, processes, and acts on data streams without being summoned - generating outputs that appear in the user's environment as relevant context, completed tasks, or proactive recommendations. Examples include AI that sweeps news feeds to surface relevant competitive intelligence in a marketing dashboard, systems that pre-populate spreadsheets from incoming emails, and tools like Aryn that process documents and deliver structured outputs to downstream systems without the user ever opening an AI interface. The contrast is with reactive AI (chatbots, on-demand tools) that require the user to initiate each interaction.

Coordination that is simply present in the workflow rather than enforced through meetings or dashboards.

An AI system that passively listens, captures, and processes clinical encounters without requiring active input, then automates downstream documentation, coding, and operational tasks.

Capturing the consultation naturally without the clinician changing behavior, no dictation or commands required.

AI that listens during a visit and drafts the clinical note so the physician faces the patient, not the screen.

Autonomous Mobile Robot, a robot that moves goods or itself through facilities.

The fear processing center of the brain, referenced as the biological source of scarcity thinking.

A fixed still image used to lock the visual start, middle, or end of a scene. The video model bridges between anchor frames rather than inventing its own transitions.

A room engineered to absorb nearly all sound. Higgins spent time in one during his PhD and could hear himself breathing through his own nose.

A global early-stage investor that backed Levelheaded.

Average Order Value.

Application Programming Interface, the connection point that lets tools talk to each other.

Limiting what an agent can access and do via controlled interfaces.

A private password that lets one piece of software use another's services.

The file format for installing Android apps.

An Apple iOS feature that loads a native app experience on a user's phone from a single link — no App Store download, no TestFlight, no invite code required. App Clips can now be up to 100MB in size.

The process of delivering a full native app experience through an App Clip container, bypassing the traditional App Store distribution model entirely.

The technical and policy requirements Apple and Google impose on apps before they can be listed in the App Store or Play Store: account deletion functionality, data privacy disclosures, restricted API usage policies, content guidelines, and in-app purchase rules. Apps that fail these requirements are rejected and must be revised before resubmission, often adding days or weeks to launch timelines. Woz pre-bakes these requirements into every generated app by default — the result is passing App Store review in an average of one to two rejections, which outperforms many hand-built apps.

A request for a higher court to review a lower court's decision.

A clinician-completed summary used when an electronic record cannot be retrieved, common in insurance underwriting.

The Advanced Research Projects Agency for Health, the innovation arm of Health and Human Services, modeled on DARPA.

The output produced by deploying a lens against information: a brief, deck, code block, or analysis. AI makes artifacts abundant; lenses remain scarce.

A network of AI agents that share information and coordinate their actions so they behave like a single collective system.

A cultural correction triggered not by better technology or stronger arguments, but by widespread felt wrongness about the dominant creative form. Historically, renaissances begin when culture moves — before the statistics confirm it and before the industry acknowledges it. The emotional signal precedes the data signal.

Third Way's EHR-integrated patient experience layer for scheduling, intake, reminders, and surveys.

The middle path between passive communication (avoiding conflict, not expressing needs, hoping others will infer what you want) and aggressive communication (attacking the person, using blame, creating psychological threat). Assertive communication is firm, direct, explicit about needs and consequences, and respectful of the other person's perspective simultaneously. The NVC framework is one structured approach for achieving assertive communication in high-stakes conversations. Most technically skilled founders default to one extreme — either conflict avoidance or bluntness — and miss the assertive middle path.

A platform setting that controls how directly and persistently an AI voice agent drives toward the next stage of the sales pipeline.

The smallest unit of implementation work an AI coding agent can execute in a single turn without requiring additional context, clarification, or mid-task correction. An atomic task has a defined start state (what exists in the codebase), a defined end state (what should exist after), and clear acceptance criteria (tests pass, linting clean). The goal of requirement decomposition in AI-native development is to produce tasks atomic enough that the agent completes each one before context drift introduces errors. Tasks that require mid-execution judgment calls are not yet atomic — they need further decomposition.

The database of record for candidates and stages. Zero Hiring supplies one if the client does not have one.

A psychology framework describing how people bond and react to closeness or distance (secure, anxious, avoidant).

An ecosystem where human focus is monetized and competed for.

A technique that lets a neural network focus on the most relevant parts of its input when producing each piece of output. It is the core idea behind transformers.

Greg Eisenberg's product development framework: build an audience around a shared interest first, let it self-select into a community with shared identity and recurring pain, then build the product that solves the community's specific problem. Solves cold-start on both distribution (the community IS the launch audience) and validation (the product spec is the community's recurring frustration, not a hypothetical market). ID345 (vibe coding meetup → community of 300+ builders) to On Demand Human (product solving that community's 80-90% wall problem) is the canonical live example.

When your audience's rewards pull you toward a more extreme or polarized position than you actually hold.

A recorded history of actions taken by a system, user, or agent for debugging, accountability, and compliance.

The ability for someone else to inspect and verify how an AI system behaved.

A tamper-evident record of everything an AI agent accessed and did during a task — required for debugging when agents make mistakes and for compliance in regulated enterprises.

A transparent log of how an AI system arrived at its conclusions, including citations and reasoning steps. Required for legal AI deployments where attorneys must be able to verify and defend every recommendation.

Owning what you hide so it loses its power over your leadership. Not performative sharing — owning the specific fear that is driving your default behavior.

Verifying who a user is when they log in. One of the most commonly botched parts of AI-generated apps — and the reason most vibe-coded prototypes are not safe to put in front of real users.

A specific schema markup type that signals who wrote the content. Boosts credibility signals for Google and LLMs evaluating expertise, authority, and trustworthiness (E-E-A-T).

A decentralized-finance mechanism that creates liquidity by pooling assets and matching trades algorithmically rather than through a traditional order book.

Technology carrying out specific tasks according to rules a human has defined. Unlike an agent, it follows the instructions rather than deciding how to reach a goal.

An AI system that sources, matches, and closes sponsorship and partnership deals for creators without a human sales team in the loop.

A persistent simulated world where agents generate events without human control.

Microsoft's new category of always-on agents with their own identity, announced at Build 2026, beginning with a product called Scout.

The practice of systematically increasing the average dollar amount per customer transaction through upselling, cross-selling, or bundling. In restaurant contexts, AOV optimization traditionally relies on staff training to prompt upsell suggestions (adding dessert, upgrading to a larger size, suggesting add-ons) - but human staff execute this inconsistently, typically upselling on 5-8% of orders. AI voice agents upsell on 100% of orders, because the prompt is built into the conversation logic and never gets skipped due to time pressure, distraction, or forgetfulness. Christian Wiens cites this as one of three key financial mechanisms Loman delivers: even if only 30% of customers accept an upsell, applying the prompt to every single transaction produces a measurable lift in revenue per call versus the human baseline.

Packaging consumer insight as a product sold to brands.

Business to government sales, notorious for long payment terms.

Grobler's term for permission harvested through platform terms of service that users accepted without understanding the likeness implications.

Rob Collie's term for company context an agent can look up on demand, such as go-to-market strategy or competitive differentiation.

Links from other websites pointing to yours. The primary currency of SEO credibility. Earned through PR, podcasts, press releases, and industry media — not purchased in bulk.

How neural networks learn from mistakes: the error is traced backward through the network to work out which connections to adjust.

A viral YouTube horror series by Kane Parsons, adapted into a 2026 A24 film, known for its unsettling "liminal space" office corridors.

The accounting process of matching bank records to internal books — an early target for AI replacement.

The neurological process by which repeated behaviors - including writing - become habituated in the brain's habit center, making constructive responses automatic over time.

A model that maps how variables influence each other using probabilities, useful for reasoning about causes and likely outcomes.

A screenwriting outline that maps the key story moments in order.

Identifying users by their unique behavior patterns — location routine, typing cadence, device usage — rather than physical traits or credentials.

A clinical category covering conditions where behavior is the primary symptom and the underlying cause is not a discrete pathogen: anxiety, depression, PTSD, schizophrenia, bipolar disorder, substance use disorder, and related conditions.

Tools and practices for turning company data into reports and insights.

When an AI system produces results that systematically favor or disadvantage certain groups, often because of skewed training data.

Datasets too large or complex for traditional software to handle, which AI can analyze to find patterns and trends.

The four largest accounting firms: PwC, EY, Deloitte, KPMG.

The legacy law pricing model Logan is trying to retire.

SpotitEarly's term for a platform where a biological sensor sits at the front of the pipeline and machine learning handles interpretation and verification.

Non-contact measurement using a device's speaker and microphone to detect breathing and movement across a room, no wearable required.

How far the damage spreads when an agent fails. Higgins's core pricing input.

Unknown weaknesses or missing validation steps.

An AI-powered mobile app builder that generates fully native iOS, Android, and web apps from a text or voice prompt. Each app includes authentication, a real-time backend, and instant sharing via App Clips.

Professional financial data platforms that cost thousands per seat.

Kindling Solutions's diagnostic engagement that maps how work actually moves, finds the constraint, and defines what to build first. Listed at 21 days and $5K.

The Buddhist commitment to delay one's own enlightenment until all sentient beings are liberated.

A scan that measures lean mass, fat mass, and bone density separately — giving a far more complete picture of health than scale weight alone.

The system that takes direct reservations on a hotel's own site.

A calendar link — Calendly being the canonical example — that lets others self-schedule a meeting based on shared availability. Will argues this approach is commoditized and inferior to assistant-style scheduling.

Bookings are contracts signed, run rate annualizes a recent period, ARR is recurring subscription revenue. They are not interchangeable.

Babysitting an off-the-shelf model through a task by feeding it context verbally, in real time, every session.

Estimating potential revenue based on customer count and pricing.

The evolution of BPO with technology layered on top — automation, AI, and analytics rather than just labor arbitrage.

A strategic outcome in which a country or region retains its educated and skilled workforce by creating high-quality employment opportunities locally, reducing the economic pressure to emigrate for better opportunities. In the Philippines, brain drain has historically forced talented professionals to leave for higher-paying jobs in the US, Europe, and the Middle East. Offshore staffing models like Penbrothers contribute to reversal by connecting Filipino talent with global companies without requiring the worker to leave - creating competitive salaries and career opportunities that were previously only available abroad. Nicolas identifies this as a core part of the Penbrothers mission, not just a market strategy.

Emberos's proprietary mapping of how a brand's entities, products, narratives, and digital assets relate to each other across the internet, enabling predictive visibility modeling.

The accumulated context of a brand's decisions. Usually stored in people's heads, which is why it disappears during turnover.

Large-scale field crops like corn, soybeans, and wheat, distinct from vegetable and specialty crop production.

Letting Meta's algorithm find the audience instead of hand-defining a narrow one. Now the platform default.

The pattern where AI applied to a dysfunctional system speeds up the dysfunction rather than correcting it. AI amplifies the system it sits inside — a broken system running faster is still a broken system, just more visibly and consequentially so.

A regulated entity licensed to transact securities on behalf of clients. Fairmint is working to add this registration.

An existing real world environment not designed for robots.

An existing industrial or commercial space built before robots were considered. Deploying robots in brownfield spaces requires adaptable navigation (like Visual SLAM) because the environment wasn't designed for them — no floor markers, no infrastructure changes.

Land previously used for industry where environmental disturbance already occurred.

Jason's term for the decision framework that determines whether a given idea should be built at all.

A founder and startup growth strategy in which a company shares its product development, business metrics, strategic thinking, and operational learnings openly - typically on social platforms like LinkedIn, X (formerly Twitter), or in public forums. The strategy serves multiple purposes simultaneously: it builds brand recognition ahead of the sales cycle, creates community around the founder's perspective, generates inbound leads from buyers who self-select based on resonance with the shared thinking, and establishes the founder as an authoritative voice in their category. Dan Bladen credits Kadence's LinkedIn presence with enabling enterprise deals that would have been inaccessible through cold outreach - including replacing a 10-year competitor in nine weeks after the buyer had been following Kadence's content.

Software plus the governance, permissions, security, and process needed for an organization to rely on it.

The chronic emotional and physical exhaustion from sustained high-stress caregiving, compounded by poor communication culture and inadequate emotional preparation. A leading cause of attrition in the nursing profession.

A future-facing use case where AI could identify early signs of physical or emotional decline through ongoing capture.

The rules and decisions that define how a specific company actually operates. The part AI cannot infer for you.

A software architecture philosophy that starts from business requirements, user trust, compliance constraints, and production quality standards — then designs the system to meet them. Contrasts with technology-forward development, which starts from available tools and finds appropriate use cases. At Woz, this means making CTO-level architectural decisions before any code is generated, encoding security guardrails as system defaults, and removing options that would allow founders to unknowingly introduce vulnerabilities. The analogy: putting the safe scissors out and locking up the adult scissors before the toddler enters the room.

An architecture pattern where enterprise customers plug their own AI vendor API key into a product so data flows through their existing legal agreements rather than a vendor's.

Letting a coding agent run without step-by-step approvals.

Computer-Aided Design software where 3D modeling of physical parts and assemblies happens. Autodesk, PTC, and Siemens are the dominant incumbents.

Compound annual growth rate; lithium demand grew roughly 30% annually over the past decade per Jeremy.

Collective Health's AI suite built with Google Cloud, spanning a member concierge, an agent copilot, and an enrollment tool.

When two events are scheduled on top of each other. Workmate proactively flags these before they become a problem.

A probability estimate that accurately reflects the true likelihood of an outcome over many predictions: if you say something has a 70% chance, it should materialize roughly 70% of the time. Calibration is a distinct measure from raw accuracy - a model can win most of its bets while being badly miscalibrated (e.g., stating 95% confidence on events that only happen 60% of the time). LightningRod measured its forecasting model on Polymarket using calibration as a primary metric, not just profit or win rate. Ben Turtle uses Polymarket bets as a forcing function for calibration training: putting money down on a prediction and seeing the outcome is the fastest way to build an accurate internal probability estimator.

Bodyweight-based strength training emphasizing control and mobility.

The optical device that reshaped painting, and her historical analogy for what AI is doing to design.

Injecting fake video into a device as if it came from the live camera. Incognia detects the spoof rather than the fake video itself.

Canada invests approximately $53 billion annually in R&D — roughly $40 billion directly into universities and research institutions. The University of Toronto is credited as a birthplace of modern deep learning (Geoffrey Hinton's research, which contributed to TensorFlow and became the foundation of modern LLMs). But Canada commercializes only a small fraction of this IP compared to US peers; a single top US university may outpace the entire Canadian system. In the AI era, as software-as-moat disappears, deep IP becomes the only defensible competitive edge — making this commercialization gap strategically urgent.

The outer limit of what a person or team can produce at a given level of staffing and skill. AI extends this boundary — enabling smaller teams to produce work previously requiring larger ones. Section operates with 32 people doing the work of 45. For founders, this changes the build vs. hire calculation: always audit your AI-extended capability before adding headcount.

The gap between what frontier models can already do and what they're actually being used for.

The ratio of revenue generated to capital invested, as redefined by the AI-driven collapse in execution costs. Pre-AI, reaching $1 million ARR typically required $2-5 million in capital (engineering salaries, infrastructure, sales team). AI-native companies designed around the outcome-as-a-service model can now target $1 million ARR on less than $1 million in capital - a multiple that was structurally impossible before AI collapsed the labor cost of software execution. Infinity Constellation's portfolio benchmark: multiple companies reaching $1M ARR with teams of five or fewer in their first year. Capital efficiency in the AI era is therefore not just a financial metric but a proof point for the underlying model - if AI leverage is real, the unit economics should be visible in the revenue-to-capital ratio from the first year of operation.

A business that generates significant revenue per dollar of investor capital raised — the holy grail for growth-stage investors.

The process of raising and issuing new equity to fund a company. Fairmint automates and lowers the cost of this process.

The business development process of pursuing a specific government opportunity from identification through proposal submission.

A multi-slide social post that walks through an idea slide by slide.

Base salary plus signing bonus plus any guaranteed cash. Equity and discretionary bonus are excluded. This is the number Zero Hiring's 6% is calculated against.

Matching incoming bank payments to the specific invoices they pay off.

A therapy approach that changes behavior by examining and reframing thought patterns.

Cognitive behavioural therapy for insomnia. The first-line non-drug treatment, and one of the intervention categories Sleep.ai has studied.

A letter demanding someone stop infringing, often used when a competitor copies your work.

A natural limestone sinkhole filled with groundwater, common across the Yucatán. Fresh water access is a major reason the Maya thrived in the region.

A trained neutral who guides mediations. At Levelheaded, every agreement is approved by a certified mediator.

An angled edge cut into a part instead of a sharp corner. A chamfer is almost always cheaper to manufacture than a sharp corner or a radius, but many designers over-tolerance these features unnecessarily.

An AI agent built on top of an existing IP's lore, voice, and personality.

Keeping a generated character’s face, age, voice, and look the same across shots.

A program that holds a conversation with people through text or voice.

OpenAI's health experience, launched to US users 18 and older on July 23, 2026, connecting Apple Health and supported medical records.

The marketplace bootstrapping challenge where supply won't join without demand and demand won't come without supply. Typically solved by subsidizing one side — usually supply — early on.

The white clay mined in Cornwall for 275 years; the Trelavour lithium project sits in a former china clay pit.

Persistent difficulty falling or staying asleep. Colin cites 800 million people worldwide, mostly untreated, though global prevalence estimates vary widely by definition.

LightningRod AI's core training method: given all data up to timestamp A in a corpus, predict what will happen at timestamp B. The model is trained iteratively across an entire dataset of chronologically ordered real-world information. The 'label' is simply what actually happened next - no human annotation required. This approach forces the AI to learn causal relationships rather than surface patterns: to get good at predicting timestamp B, the model must understand what factors in timestamp A's data actually drive outcomes. Ben Turtle calls this the mechanism humans use to build world models - we move through time, predict what comes next, feel surprise when we're wrong, and update our beliefs.

Timing a medical treatment to the patient's circadian rhythm. Evidence is genuinely mixed.

Your biological bias toward mornings or evenings, morning bird, night owl, or somewhere between. Colin's argument is that prescribing 6am workouts to a night owl is prescribing failure.

The rate at which customers cancel.

Privacy professional certifications from IAPP covering European law, program management, and technology implementation respectively.

The roughly 24-hour internal clock, entrained primarily by light exposure, that governs sleep timing, hormone release, and metabolism.

Cybersecurity and Infrastructure Security Agency.

Chief information security officer.

How often your specific URL is referenced inside an AI answer — a stronger signal than a mention alone.

Returning answers tied to specific document sources.

The process of reviewing and deciding whether and how much a medical claim gets paid.

The socioeconomic and cultural gap that keeps software engineers from discovering and solving unglamorous problems in traditional industries. The founders at elite technical schools build for space and quantum because that is who they know — not because those problems are more valuable. The gap between what software engineers know and what lower-middle-market operators know is where the real AI arbitrage lives.

An AI assistant that reflects the broader trend of AI tools showing up across many software development environments.

Anthropic's agentic knowledge-work app for non-developers. Rob Collie's recommended entry point for people who don't want to touch code.

The desktop Claude application, which can connect to external tools and data through MCP.

Anthropic's education product, launched July 14, 2026, free for verified US K-12 educators, connected to state standards via Learning Commons.

Anthropic's partner program, tied to the Anthropic Partner Academy, which gates access to certification exams and starts free at the Registered level.

Text-based interface for interacting with software. Tim's transition from not knowing what a CLI was to running multiple Claude Code terminals daily marks the operator's full conversion.

Clinical Laboratory Improvement Amendments, a federal certification program covering labs. A CLIA registered lab can develop and run a test in-house and sell the result, separate from FDA device approval.

The churn cause that occurs when a customer signs up with incorrect expectations about what the product delivers. At Stone Systems: contractors who believe $297 buys direct lead generation (spend → get leads) rather than a software platform that builds online presence and operational efficiency. Mismatched expectations are set before the sale and corrected only through better marketing copy, cleaner positioning, and sales qualification — not through product changes.

One-on-one relationship selling where associates use customer data to drive personalized outreach and repeat purchases.

AI-powered transcription and note generation that listens to provider-patient conversations and produces structured clinical notes, reducing administrative burden and freeing clinicians to focus on the patient rather than the keyboard.

An AI-generated probabilistic assessment of a patient's health status or risk level, derived from multiple data inputs. Not a diagnosis — a statistically grounded estimate ('the likelihood of this person having hypertension is above population average') that requires clinician review before any action is taken. Axenya generates 95 million clinical inferences per month across its monitored population. The key distinction: a clinical inference routes a patient toward care; a diagnosis determines what care they receive. The former is automatable at scale; the latter requires physician judgment.

Proprietary software whose code is kept private, so the public can't see, modify, or reuse it.

Precision material cutting, usually metal but also plastic and wood, using computer-controlled machine tools. Tolerancing and chamfers in the design directly drive CNC cost.

The practice of partnering with other brands or individuals to share guest lists, split costs, and reduce hosting anxiety. The list share is often the primary value.

Shared residential housing with private rooms and communal amenities. The model at Ollie.

Composition of Foods Integrated Dataset, the UK government's official food nutrition database, used alongside USDA data in Qalzy's 2.9 million item database.

Automation that uses NLP and ML to handle unstructured data and judgment-based tasks, sitting between RPA and full agentic systems.

The behavioral state in which reliance on AI tools becomes so habitual that working without them creates cognitive discomfort. Greg Shove's term for the addiction-like quality of AI subscriptions — not sensationalist, but behavioral-economics precise. His remedy: build AI-free time into your routine to maintain the independent cognitive capacity that makes you valuable with or without the tool.

The mental effort required to use a piece of software. High cognitive load - unlabeled icons, ambiguous navigation, unpredictable behavior - is the primary source of user frustration in enterprise tools.

The mental strain that occurs when incoming information exceeds a person's processing capacity. An emerging mental health risk for founders and operators navigating the pace of the AI era.

Cost of goods sold. The direct cost of producing what you sell.

A legal contract for unmarried couples who live together, functioning like a prenup under common-law family law regimes.

What percentage of users who joined in a given period are still active later. The QR code app held 20% at year five, which is exceptional.

Geoff Gibbins' framework for measuring the quality of human-AI collaboration across three dimensions: (1) results — the actual impact of the collaboration, not just activity or usage; (2) relationship quality — how well humans engage with AI as a collaborator, including critical thinking, good question formation, and not passively accepting first outputs; (3) resilience — whether the human is developing domain expertise through the collaboration or outsourcing judgment and atrophying. Most enterprise AI measurement stops at usage metrics; the CRI measures whether collaboration is making people genuinely better.

A curated subset of files indexed for AI interrogation.

When a once-valuable capability becomes cheap and standard, collapsing its margin.

The 70% of every new application that repeats across products: authentication, user management, CRUD operations, data pipelines, role-based access control, notifications. Not innovative at any point in software history. Every dev shop built internal boilerplates for these components. The insight: competitive advantage in software has never come from the commodity layer. In the AI era, this is more true than ever — any decent CRM can be built in a day. Abstract and automate the 70%; focus human effort on the differentiating 30%.

A couple treated as legally married under certain jurisdictions due to cohabitation, without a formal marriage ceremony.

Structured analysis of your market position against named competitors, including adjacent entrants you aren't currently watching.

A defensible advantage such as IP, proprietary data, or network effects.

Software that turns a specification or source code into a working program.

In this episode, the percentage of users still logging food after a given period. Kostas cites roughly 60% higher compliance for scale users.

Combining multiple images or layers into one frame, often in Photoshop.

A capability that allows AI agents to operate a computer directly, clicking buttons and navigating interfaces the way a human would. Will's team uses it to manage paid advertising.

The branch of AI that lets computers interpret images and video, like recognizing objects, faces, or text.

Exploring many design and manufacturing paths in parallel during the early design phase, rather than sequentially. Requires fast tooling — historically impractical for large assemblies.

A human psychology term for the way memory reconstructs rather than records — people create plausible narratives around incomplete memories with no intent to deceive. Geoff draws the parallel to AI hallucinations: both are a system constructing a coherent story from incomplete information. The observation cuts both ways: AI hallucinations aren't as alien as they seem (human memory does the same thing), and human testimony is far less reliable than assumed (well-established in criminal justice research). Working closely with AI has become a useful mirror on human cognition.

A technique for improving LLM reliability: after receiving an answer or solution, ask the model to rate its confidence level in what it just provided. A low rating (e.g., 'I'm about 70% confident') typically triggers the model to reconsider, surface suppressed uncertainty, and work toward a more validated answer before you act on it. Useful for code solutions, factual claims, and strategic recommendations. Ran uses this routinely: 'What's your confidence level in the solution you just created?' — the model often responds by catching its own errors.

Technology whose primary value is facilitating human connection, emotional processing, or felt experience — rather than productivity, content creation, or time savings. Deep Gem Interactive's generative music platform is connection technology: it creates emotional artifacts specific to a person and a relationship, not optimized for distribution or engagement.

The third and highest stage of Neuro Spiritual Sovereignty. Acting in alignment with a higher mission even when intrinsic motivation is absent.

The average analyst forecast for a metric. FNS compares results against FactSet consensus.

Data supplied with explicit, informed user permission. The regulatory and commercial bar that frontier AI labs increasingly demand.

The rules, values, and ethical principles baked into each agent at instantiation. The foundational constraint set that shapes all downstream behavior before any prompt is processed.

A pricing model in which customers pay based on actual usage rather than seats or licenses. Forces the vendor to care about adoption velocity and ongoing engagement — because revenue scales with real usage, not contract value. When Knapsack moved to consumption pricing, time-to-value became the primary commercial driver: the faster a customer gets value, the faster they consume more, the faster revenue grows. Aligns vendor and customer incentives in a way seat-based pricing does not.

The many formats, short video, long post, thread, spun from one core idea.

A system that publishes useful content on a schedule to attract clients.

The core argument of a piece, captured in the creator's own words, that all formats are built from.

A named configuration of rules governing how an AI generation system behaves in a specific mode — for example, 'creative exploration' vs. 'compliance review.' The same underlying model produces reliably different outputs depending on which archetype is active. In Knapsack's Intelligent Product Engine, archetypes are one of three components (alongside real-time artifact monitoring and a rules engine) that make AI generation predictable and enterprise-appropriate rather than open-ended and unpredictable.

The practice of providing an AI coding agent with accurate documentation of the existing codebase before asking it to generate or modify code. Without grounding, agents operate on generic code patterns and produce architecturally inconsistent output that may be technically correct in isolation but incompatible with the existing system. BrainGrid generates five grounding documents from the actual repository — architecture overview, key workflows, data structures, directory structure, API patterns — so the agent extends what's there rather than inventing new patterns.

A hallucination that occurs not from false information, but from failure to account for relationships between documents, where one document modifies, contradicts, or conditionally applies to another.

How much input, such as text, documents, or conversation history, an AI model can take into account at one time.

The practice of measuring health variables throughout a patient's daily life — during sleep, work, exercise, stress — rather than episodically at clinical visits. Blood pressure measured once at a doctor's office reflects a single moment and is subject to white-coat effect; continuous monitoring captures the longitudinal patterns (variability, trends, activity correlations) that predict future chronic disease events. Chronic disease management requires this longitudinal picture because chronic diseases progress silently between visit-based measurements and have no early symptoms to motivate clinical visits.

A care model in which patient data is collected between appointments rather than only during visits. Enables longitudinal trending, earlier detection of deterioration, and proactive intervention before a crisis requires emergency care.

Structured fields attached to a contract document - vendor name, value, expiration date, key terms. AI procurement tools can surface insights even when metadata is incomplete by interrogating the unstructured contract text directly.

Structured alternation between high heat and near-freezing immersion.

A training technique where a model learns by comparing similar and dissimilar examples, figuring out what makes things alike or different.

The umbrella category covering greenhouses, indoor farms, and vertical farms where light, climate, water, and nutrients are managed.

A Bonds feature where AI plays your partner so you can rehearse a hard talk and get scored feedback.

A novel AI interaction modality pioneered by Quarterzip: a real-time AI voice agent that consumes the user's screen share as its primary visual input, enabling it to provide contextual, goal-oriented onboarding guidance based on what the user actually sees — not what they describe in text. Unlike traditional chatbots that respond to text queries, conversational screen share gives the AI the same spatial awareness a human trainer sitting beside the user would have, allowing it to recognize page state, workflow position, and error conditions in real time.

A backend-as-a-service platform providing real-time data sync between devices, full type safety, and serverless functions. Bloom uses Convex as its backend, enabling live multi-device sync from a single prompt.

The cumulative time cost of all the back and forth required to schedule meetings, resolve conflicts, and reschedule when things change.

The infrastructure that handles all the in-between work in a knowledge workflow. Workmate is building the coordination layer for meetings.

Seats for Microsoft 365 Copilot, which Santosh says are a weak proxy for AI ROI.

Snowflake's agent-based coding offering — the incumbent enterprise data platform's answer to AI-driven engineering automation.

The stress hormone response repeated alarms may trigger. The mechanism behind the case against serial snoozing.

Pre-existing commercial products adapted for government use. Increasingly preferred over custom-built solutions as the government pushes for faster, cheaper innovation.

Certified Public Accountant.

The deliberate limits within which a creator operates that make human decisions meaningful. Without constraints, creativity is frictionless and produces chaos. With constraints, every choice matters — and the output feels earned. The best creative tools are defined as much by what they do not let you do as by what they enable.

When an ad's performance decays from overexposure. Cost per acquisition rising over consecutive days is the tell.

A single operating system for a brand's creative work, holding assets, rules, and decisions in one place rather than across scattered vendors and files.

The ability of an authentication credential to be copied, transmitted, stored on a new device, or used in a location other than where it was created. Passwords, API tokens, session cookies, TOTP codes, and QR codes are all movable credentials - they can be phished, stolen from memory, captured in transit, or replicated by an attacker. Jasson Casey's insight: credential movement is the root cause of approximately 80% of security incidents (per multiple annual threat reports from CrowdStrike, Mandiant, and Verizon DBIR). Device-bound credentials eliminate movement by design: the private key is stored in a hardware enclave and can never leave. If a credential cannot move, it cannot be stolen - and the dominant attack chain collapses.

An inverted prediction market where lenses compete against other lenses, generating a real-time signal of your judgment's value to allocators, employers, and clients.

The high-stakes engineering gate where a design gets locked toward production. Everything after this point gets radically more expensive to change.

Minerals essential to modern technology and defense where supply is at risk, including lithium, rare earths, tin, and tungsten.

Customer Relationship Management software; the system of record for customer data and interactions.

Raising capital from retail investors, regulated in the UK by the Financial Conduct Authority; Cornish Lithium used Crowdcube across multiple rounds.

The emergent quality indicator produced when many users independently mark moments in content they find most valuable — and the aggregate frequency of those marks per content unit becomes a discovery layer. In Snipd, snip frequency per podcast episode is the signal: episodes with high snip density contain demonstrably more insight per minute than low-density episodes. This is fundamentally different from download counts (show-level popularity) or editorial recommendations (one person's opinion). It is episode-level, content-quality-driven, and derived directly from independent listener judgment about specific moments.

The one thing you want a visitor to do next, like apply.

Internet-connected television, the ad channel behind Tal's VFR.ai.

A structured process for preparing both the client and the offshore employee to work together effectively across cultural differences. Distinct from job onboarding (learning the role), cultural onboarding addresses communication styles, feedback norms, decision-making expectations, and the specific cultural characteristics of the offshore team's country. Nicolas describes this as a core part of Penbrothers' quality investment: a new client hiring Filipino talent for the first time receives guidance on how Filipino workplace culture differs from their home culture, what communication patterns to expect, and how to adapt management behaviors to get the best outcomes. Without it, the leading cause of offshore placement failure is not talent quality - it is cultural misalignment on both sides.

An ancient system of writing developed by the Sumerians, made by pressing a reed stylus into clay tablets. Referenced in conversation about the Babylonian origins of the Noah's Ark flood myth.

Todd's mnemonic for target companies: Complex, Urgent, or Administrative.

Agentic Healthcare's proprietary network of connections to health information exchanges, with coverage for more than 128 million individuals.

The structured practice of interviewing potential customers to uncover their actual pain points, needs, and behaviors — without leading them toward your solution. The goal is to be surprised. If every interview confirms what you already believed, you were not asking challenging enough questions.

The signal that users are actively seeking out and recommending a product rather than being pushed toward it through marketing or sales. Ali's target signal during pre-PMF iteration: sprint-on-sprint, are customers pulling the product out of your hands and telling five other people to use it? The presence of pull — not adoption driven by founder sales effort — is the leading indicator that product-market fit is being approached.

Directors and officers liability. Protects leadership from claims about their decisions. Now carrying absolute AI exclusions at some carriers.

Germany, Austria, and Switzerland.

A governance structure in which decisions are made collectively by members using transparent, typically blockchain-based voting mechanisms. Mike advocates for DAOs as the infrastructure for hyper-local charitable giving (his Donaton concept): communities vote on their biggest local problems, and funding is algorithmically allocated based on scoring rather than centralized NGO decision-making. The argument: global NGOs are good at raising money but bad at finding who needs it most; local communities can vote on that with far greater precision.

The Defense Advanced Research Projects Agency, the innovation arm of the Department of Defense.

A self-contained interactive interface — often React-based — generated by an agent to visualize and interact with data without requiring a human developer to build it.

Expanding a training dataset by creating modified copies of existing data, like flipped images or reworded sentences, instead of collecting new data.

Enterprise software that indexes an organization's data so teams can find, govern, and trust it. The category data.world built in.

A curated subset of a user's files — PDFs, documents, images, voice memos — that is analyzed once by a local AI system and made queryable without any cloud access. The unit of organization for privacy-first AI interaction with personal or proprietary data.

Each use generates data that improves the product, which drives more use and more data.

The tendency of applications and workloads to cluster around large data stores.

Tagging documents so a model learns what matters in them.

A centralized repository for structured and unstructured data. Not an Excel file, no matter what the filename says.

Finding patterns and trends in large datasets using statistics and machine learning.

How information is structured and related so a system can query and trust it.

An attack where someone slips corrupted, altered, or mislabeled examples into a model's training data to make it behave badly or carry a hidden backdoor.

The verified, cryptographically attested record of where data came from - which device produced it, which identity authorized it, through which workflow it passed - as a security property rather than just a metadata label. In the AI era, data provenance is the architecturally correct answer to synthetic media and deepfakes: instead of trying to detect whether content was AI-generated (an arms race that detectors will lose), you attest the chain of custody from origin to recipient. Jasson Casey's prediction: device-bound credentials will extend beyond authenticating humans to attesting the provenance of data produced by cameras, microphones, sensors, and implants. The camera on a phone will have a passkey; the microphone will have a passkey. Anything that generates consequential data will eventually need a way to attest its origin - not because the content cannot be faked, but because provenance establishes authorization.

Where your records live — Airtable, Google Sheets, Supabase, or a REST API. Softr connects to these sources and builds the interface and logic on top without migrating the data.

Raw clinical data — labs, notes, images, wearable readings — becomes information only when it is integrated, structured, and interpreted in context. Most healthcare systems generate enormous amounts of data; very few generate actionable information. The Mercatus/Diagnostic MD architectural thesis: the value is not in capturing more data but in transforming fragmented data into structured, visual intelligence that a specialist can act on.

A centralized repository where structured data from multiple sources is stored and organized for analysis. Common platforms include Snowflake, BigQuery, and Redshift.

Stripping identifiers from health data so it can be shared. The thing TripleBlind's approach was designed to make unnecessary.

The concept that AI-generated content is increasingly training on other AI-generated content, creating a feedback loop of degraded and homogenized information. Kaelan references it as the backdrop for why original, experience-grounded content increasingly wins.

White-label capital-raising infrastructure. The issuer runs the raise on its own domain while DealMaker Securities acts as broker-dealer of record. Roughly half of all Reg A+ dollar volume in 2025.

A model that makes predictions by asking a series of yes-or-no questions about the data, branching at each step until it reaches an answer.

Booking your own GPU capacity for very large workloads, sometimes for a custom fine-tune.

Machine learning that uses neural networks with many layers to learn complex patterns directly from raw data.

An automated process that gathers sources, quotes, companies, and people around a topic and outputs a research brief.

Extended, distraction-free focus blocks reserved for cognitively demanding tasks.

AI-generated video, image, or audio convincing enough to impersonate a real person.

The brain network associated with self-referential narrative, the inner monologue, and judgment. fMRI work shows psychedelics temporarily quiet it.

In game theory, choosing to act against the other party instead of cooperating.

1950 US law letting the president direct industry for national defense, now invoked for critical minerals.

A live metric used in Wildwood's pitch scoring sheet to evaluate how protected a startup's business model is from replication. Scores have declined industry-wide as AI lowers the cost of building software.

Proprietary data, SOPs, chemical formulas, machine configurations, and operational playbooks that cannot be replicated by a competitor running the same foundational AI model. The moat that makes a niche AI product acquisition-resistant and justifies enterprise pricing. Josh Furstoss's second checkbox before starting any new company.

A deliberate strategy of publicly disclosing an invention without filing a patent, specifically to create prior art that prevents competitors from obtaining a patent on the same idea. The disclosing party waives any future patent rights but ensures no one else can obtain them either — the invention effectively enters the public domain immediately. Used when a company wants to operate freely in a technical space without the cost, time commitment, or disclosure-and-expiry constraints of a patent. Ophir describes this as one of SenseIP's emerging capabilities: helping companies identify which portions of their IP to publish defensively rather than patent.

The percentage of support requests resolved entirely by AI without escalation to a human — Unthread's primary north-star metric.

The default entity structure for startups planning to raise venture capital.

The full path from acquiring a customer through delivering the thing they bought, plus the internal functions supporting it.

A generalist venture capital fund based in Denver that invests across multiple industries at early stages, led by a thesis centered on founder DNA.

Publishing code so the site is live on the internet.

When a vendor retires a tool or model, forcing users to migrate.

A video and podcast editor that lets you edit media by editing the transcript.

An early-access cohort of customers who help shape a product in exchange for founder-level pricing and direct access to the builder. Chime's program caps at 25 seats at $39/month locked in for life.

Protects how something looks; typically easier to interpret and litigate.

A central system of record for all building blocks of a digital product — 'a well-organized bucket of Lego bricks' containing every component, variant, rule, and constraint. More than a component library: a design system encodes the rules governing how elements connect, making it the substrate for consistent product experiences and enterprise-appropriate AI generation. Without a structured design system providing context, AI-generated UI is a prototype tool; with one, it becomes a production capability.

The percentage of patients who stop taking a medication, often due to side effects or cost. GLP-1 desistance rates in standard protocols run 50–70%.

A workplace strategy in which employees do not have permanently assigned desks but instead reserve a workspace for the specific times they plan to be in the office. Also called hot desking, though the two terms have subtle distinctions: hot desking typically refers to first-come-first-served open seating, while desk hoteling involves advance reservation. Both models are designed to accommodate hybrid workforces where not everyone is in the office on the same day, enabling companies to support a headcount significantly larger than their physical desk count - often a 2:1 to 4:1 person-to-desk ratio. Kadence's core functionality includes desk hoteling management, ensuring teams who need to overlap get adjacent space and preventing any day from being oversubscribed.

Produces the same output for the same input every time.

Agent-driven automation with predictable, governable outcomes, designed for regulated enterprise use cases where explainability and auditability matter.

A software system that ingests structured data inputs and produces consistent, rule-based outputs, as opposed to generative or probabilistic AI responses.

A use of AI that converts natural language into a fixed, repeatable result — for example, translating a user question into a precise database query that always returns the same answer for the same input.

A software system that produces the same output for the same input every time, with no probabilistic variation. Succession Wills uses a deterministic system to generate will language, ensuring no AI improvises legal text.

Tasks that must be exactly right every time (medical, materials) versus tasks where occasional error is tolerable. The skill is separating the two.

Deterministic: run the same computation twice, get the same answer twice. Stochastic: probabilistic output that varies. Hardware demands deterministic. LLMs are stochastic — which is why Encube isolates them from the core math.

A profile built from a device's characteristics used to track it over time and flag fraud.

The security health state of a device at the moment of authentication, used as an input to access control decisions. A device posture check can verify that specific security controls are active (EDR software running, OS updated, disk encrypted, no jailbreak detected, specific executable files signed by expected OEMs) and append that verification to the authentication assertion. Posture checks can occur silently - without interrupting the user - or can require explicit user interaction for high-sensitivity operations. Beyond Identity's policy engine lets customers tune their posture requirements on a spectrum: from a simple 'verify not jailbroken' for a retail purchasing app, to 'verify every process in memory and the loader chain' for national security customers. The policy is customer-defined and reflects the organization's actual risk tolerance for specific operations.

A cryptographic key or authentication credential that is generated inside a hardware enclave on a specific device and physically cannot be copied, exported, or used on any other device. In contrast to passwords (which exist in memory and can be transmitted), device-bound credentials never enter general purpose memory - they can sign data but cannot move. The possession factor in authentication is therefore the device itself rather than a secret that can be stolen. This model underlies passkeys, Apple Pay, tap-to-pay credit cards, and Beyond Identity's authentication platform. Jasson Casey's key insight: since 80% of security incidents trace back to credentials that can move - passwords stolen, tokens copied, session cookies hijacked - device-bound credentials break the dominant attack chain at the architecture level rather than at the detection level.

The engineering phase after prototyping where a design is adapted to be producible at scale on real machines.

Generative models that learn to create images and other content by starting from random noise and removing it step by step.

Jack's term for competitive selection in a world of universally available AI tools, where distinctiveness rather than access decides who survives.

A generation that grew up with smartphone-quality cameras and expects AI-driven, personalized digital experiences as a baseline.

The legal term of art appearing across SAG-AFTRA contracts, California statute, and the NO FAKES Act for a computer-generated representation readily identifiable as a specific person.

The SAG-AFTRA 2025 Commercials Contract provision requiring clear and conspicuous written consent, with specific descriptions of intended use, for AI-generated likeness.

A software-driven, evidence-based intervention used to prevent or manage a medical condition.

The reduction in existing shareholders' ownership percentage that occurs when a company issues new shares, typically during a funding round.

Simplifying data by cutting the number of variables while keeping the information that matters most.

The Greek god of wine, festivity, and ecstasy, known as Bacchus in Rome.

Listing existing shares on an exchange without raising new capital. Shareholder to shareholder trading, no IPO roadshow.

Bookings made straight with the hotel, bypassing OTA commissions.

A decision-making principle: don't wait for a perfect destination to act. Know what directions you're not going, pick a direction you are going, and move. You can only connect the dots looking backward — skills accumulated in one domain show up as unexpected assets in another. Passion follows competence and is never the starting point. More useful than waiting for clarity you won't have until you're already in motion.

Music created without human intent, emotional investment, or constraint — the default output of generative AI music tools mass-producing content for algorithmic social feeds. Statistically ubiquitous; culturally hollow. The antithesis of music as the language of emotions.

The thesis that owning an audience or community is now more durable than owning proprietary code.

A co-founding partner brought in specifically for their ability to reach and grow a customer base. A founding-level commitment to treating distribution as a first-class company function.

Giving equity — not just commissions — to people who already have relationships, access, and trust inside your target market. Not hiring salespeople after the product is built, but recruiting co-builders who are already known inside the networks you need to reach. Josh Furstoss's first checkbox before starting any new company.

Technology for pulling lithium directly from brines and geothermal waters rather than traditional evaporation ponds.

The AI-powered capability to extract structured, actionable information from unstructured documents - reading a PDF or scanned form and identifying specific fields, values, relationships, and entities with high accuracy. Distinguished from simple text extraction (OCR) by the ability to understand context, identify relevant fields by meaning rather than position, handle varied document layouts, and maintain accuracy across documents that differ significantly in structure. Aryn's core product is document intelligence for enterprise use cases: insurance submissions, legal filings, real estate offering memos, construction estimates. The commercial value: replacing hours of manual data entry with automated extraction that feeds directly into downstream workflow systems.

The Department of Government Efficiency, which Joe cited as inspiration for recruiting young AI engineers.

Document Object Model, the live structure of a webpage that Cora manipulates without touching your code.

Compulsively consuming negative or low-value feeds.

Dr. Dhruva's term for delaying gratification while remaining open to pleasure - choosing mission-aligned action first rather than suppressing desire entirely.

Fraudulently financing the same invoice with two different lenders.

Brett Hurt's term for insights that arrive during meditation.

A structured risk review required before running processing activities likely to result in high risk to individuals' rights and freedoms.

The two AI coach personas inside Bonds.

Loftie's proximity feature that blocks distracting apps when your phone enters the bedroom.

A platform that must simultaneously attract and serve two distinct user groups whose participation depends on the other side being present. In Luxxera's case: patients and clinics.

Third Way's agent platform, named for a rock climbing move that skips a step. An umbrella for multiple agents (inbound and outbound telephony, prior auth, referrals) that work as a team.

A government database where businesses must publish invoices, making them provably legitimate.

A deal structure where part of the sale price is paid later, contingent on the founder staying and hitting targets.

Whether reported results came in above or below what analysts expected.

Mode's proprietary rewards and monetization layer, designed to sit on top of any consumer app.

Mode's smartphone, sold through Walmart, Best Buy, and other retail. Discontinued as a first-party hardware product, now the basis of an OEM licensing model.

Acquisition price expressed as a ratio of earnings. Rahi sold at roughly 7.5x.

Engineering Bill of Materials — the full, structured list of every part in a product and how they relate. The eBOM is the source of truth for what actually gets manufactured.

Any computing hardware that processes data locally without sending it to a remote server or cloud. Examples include smartphones, drones, and the Key Boy. Critical for latency-sensitive and privacy-sensitive applications.

Direct brain-wave measurement. The purest way to stage sleep, and the one nobody wants to wear to bed.

The temporary loss of the sense of a separate self, commonly reported when the default mode network goes quiet.

Electronic health record / electronic medical record, the core clinical software systems.

German/Swiss concept meaning self-responsibility: the cultural default that individuals are accountable for doing the right thing without needing explicit rules, enforcement, or social pressure to prompt them. Kevin attributes Swiss civic quality — public cleanliness, COVID policy compliance without mandates, a market that sustains premium-quality products — to this cultural norm. Contrasted with rule-based compliance cultures that respond to violations rather than cultivating internal accountability. Practically, Eigenverantwortung produces high-trust societies where quality compounds over time because the market rewards it rather than defaulting to the cheapest acceptable option.

Mayan archaeological site meaning 'black jaguar,' founded in the 3rd century BCE and inhabited through the 11th century.

A mental model for distinguishing human roles that add genuine judgment value from roles that exist only because the technology to automate them had not yet arrived. Before automatic elevators, buildings employed human operators to run elevator cars. When automatic elevators arrived, those jobs disappeared - not because the workers were incompetent but because the role itself was a placeholder for automation. Brennan applies this to hiring decisions in the AI era: any job description that could have been written identically in 2019 is a candidate for being an elevator operator role. Good headcount augments human judgment with AI leverage (one specialist doing the work of ten). Bad headcount is a human performing a task that AI can now do - a position that will be automated away as AI capabilities continue to improve, making it a capital efficiency mistake to build headcount around.

The higher-tier ElevenLabs voice cloning option that produces a clone close enough to the original to be hard to distinguish in short-form content.

Confirming a patient's active insurance coverage before a visit.

Extract, Load, Transform. The process of moving raw data into a warehouse before transforming it — one of the four highest-value use cases for data agents.

The capacity to manage, process, and integrate emotional states rather than suppress or avoid them. A primary use case for Deep Gem Interactive's generative music — using personalized sonic experience to help people move through grief, relational difficulty, or identity transitions. Also a prerequisite for sustainable founder performance.

Customer loyalty rooted in shared emotional experiences rather than transactional utility.

The state of deep self-forgiveness and other-understanding that replaces the resentment-based default patterns built during trauma.

Tyler's model running from doormat to bulldozer, with healthy reciprocal empathy in the middle. Healthy empathy means engaging with someone while still holding space for yourself.

A digital version of a patient's medical chart, used by healthcare providers for documentation and care coordination. A target data source for Ki's healthcare application — querying patient records in real time.

Automating an entire business process from input to output, rather than discrete tasks within it. The core EvoluteIQ thesis.

Engineers are system architects — they decide what to build, how components fit together, what problems the architecture needs to solve. Coders implement those specifications. The distinction matters because coders can increasingly be replaced by AI that executes specifications quickly and cheaply; architects cannot. Founders who staff with coders instead of engineers get fast execution without coherent architecture. The correct hire for a technical team is engineering judgment, not implementation capacity.

Profile, company, job-change, and signal data layered onto your records.

Using multiple foundation models in concert rather than locking into one provider's sandbox. Gives organizations auditability and reduces single-vendor capture.

Combining several models and merging their predictions to get a result that is more accurate than any single model.

The gradual degradation of product and software quality over time as cost-cutting and feature bloat compound. Coined by Cory Doctorow, used here to describe the wave of vibe-coded tools that will break constantly.

The accumulated cost of maintaining, working around, and compensating for outdated, disorganized, or incompatible data infrastructure. Analogous to software technical debt but measured at the data layer rather than the code layer. Includes the ongoing cost of maintaining legacy mainframe systems that cannot be retired without major migration, the cost of manual data cleaning and reconciliation work performed by human analysts, and the cost of enterprise AI projects that underperform because data quality cannot support the intended use case. The Wall Street Journal cites this as a $1.52 trillion annual problem in the US alone. David Carmel uses the metaphor of Victorian era plumbing: the pipes are old, they work after a fashion, but they are fundamentally incompatible with the modern infrastructure being built around them.

The April 18, 2026 executive order directing FDA to prioritize psychedelic review, instructing DEA and other agencies to reduce research restrictions, and allocating $50 million.

A goal-and-accountability framework Dave maps directly onto agent operations — roles, scorecards, rocks, and meetings all have agentic equivalents in his system.

The Traction framework pairing a big-picture founder with an operations-focused counterpart.

The highest-level planning unit in the software development lifecycle: a large body of work defining a significant product capability. An epic breaks into multiple requirements; each requirement breaks into atomic tasks. In AI-native development, epics define multi-feature projects (build a multi-tenant Pilates platform with video libraries, subscriber management, and per-studio payment systems) while requirements define individual features within that project. The hierarchy matters: agents given an epic-level prompt will attempt to solve the entire thing in one shot and fail; agents given an atomic task will succeed consistently.

Descending units of work — epics are big features, milestones are subfeatures with user stories and success metrics, tasks are small scoped changes.

Employment practices liability. Covers discrimination and related employment claims. A live battleground for AI hiring and management tools.

The idea that as AI commoditizes IQ-level tasks, emotional intelligence, taste, discernment, and the ability to make customers feel understood become the primary competitive differentiators.

A uniform unit of measurement that lets you compare cost across different product form factors built from the same input.

Email Service Provider.

Designing and deploying AI in ways that respect human values, fairness, transparency, and the wellbeing of the people it affects.

Extract, Transform, Load - a data engineering pattern describing the three-step process of (1) extracting data from source systems, (2) transforming it into a consistent, clean format, and (3) loading it into a destination system for analysis or use. ETL is the foundational workflow of enterprise data infrastructure and has been central to data engineering since the 1970s. Modern AI ETL pipelines extend the traditional pattern by using AI models in the transformation step to process unstructured inputs - documents, images, audio - that traditional ETL could not handle. Mehul Shah's background includes building AWS Glue, Amazon's flagship ETL service, and OpenSearch, a distributed search and analytics engine built on Elasticsearch.

StoriedLife’s AI biographer, offered with multiple culturally specific avatars and languages.

A Partytrick template that gives the Secret Event Planner an end-to-end structure for a specific type of gathering, covering timeline, logistics, and moment design.

Performing cardiovascular exercise, typically fifteen minutes, while breathing concentrated oxygen through a sealed mask fed by a large reservoir.

A pooled investment fund into which shareholders contribute concentrated stock positions in exchange for LP (limited partner) interests representing a proportional share of the diversified pool - without triggering a taxable sale. Exchange funds have existed in public markets for decades (Eaton Vance manages over $500 billion in exchange fund assets), allowing public company executives to diversify concentrated positions tax-free. Collective Liquidity is the first exchange fund to accept private company shares, requiring significant legal and tax structuring to enable the mechanism in the illiquid private market. The key tax advantage: a contribution to an exchange fund is not a sale, so no capital gains tax is owed at the time of exchange. The tax is deferred until the LP interest itself is eventually sold, and the deferred tax compounds over the holding period - generating approximately 2x the after-tax wealth over seven years compared to selling shares immediately and reinvesting after taxes.

The principle that execution accounts for 99% of startup outcomes and the initial idea for 1%. Ideas are the starting point — what makes them founders — but validation, iteration, delivery quality, and market timing determine whether anything becomes a business. A great idea with poor execution fails; an ordinary idea executed with relentless customer focus and iteration can succeed. The implication: invest disproportionately in execution infrastructure, not idea protection.

The person named in a will to carry out its instructions.

A message that appears when a visitor is about to leave a page.

An early style of AI that mimics a human expert using a knowledge base of facts and hand-written if-then rules.

Methods that make an AI system's decisions understandable to people, rather than leaving the model a black box.

An open-source framework built on React Native that allows developers to write one codebase and deploy to iOS, Android, and web simultaneously. Bloom uses Expo as its front-end foundation.

A hidden monitoring agent that observes all other agents without being visible to any of them. Provides an independent audit layer that no other agent can game or anticipate.

In Colorado, a state-licensed role created specifically for supervising psychedelic sessions.

Wiping a device to default settings, which generates new device identifiers and lets fraudsters dodge detection blocks. Now takes about 30 seconds on an iPhone.

The primary rulebook governing federal government purchasing above $250K. OTAs exist specifically to bypass it when speed and flexibility are priorities.

The competitive benefit of entering a validated market after initial product-market fit is demonstrated, using better tools, models, or distribution than were available to the pioneer. Distinct from being late: a fast follower enters early enough that the market is still forming, but after the founding pioneers have absorbed the discovery risk. Cal.ai is the canonical example: launched before vision AI was accurate enough to reliably identify food, it grew distribution while the models improved — and was positioned to capture compounding gains from model improvements that arrived after launch. The lesson: being in market is a prerequisite for benefiting from technology progress you don't control.

Finite Element Analysis — simulating how a design behaves physically under stress, heat, or pressure by dividing it into small elements and solving equations across each one.

The engineering and economic study that determines whether a mining project can proceed to construction.

Training a model across many devices or servers that each keep their own data, so the raw data never has to be pooled in one place.

Customer reviews that continually refine future recommendations.

A source table that collects raw inputs, such as news, transcripts, or captured ideas, before processing begins.

Jason's tool that monitors trending topics on YouTube, X, and news aggregators and surfaces story candidates.

The US federal law governing the privacy of student education records.

The average number of children born per woman. Lior cites low rates in South Korea and Japan.

When a model learns a new task from just a handful of examples, often given right in the prompt.

The HL7 data standard that lets third-party apps pull patient records out of EHR systems like Epic. Savva was built on top of this standard, and Stephen's previous company Protocol First was the first FHIR app approved for FDA clinical trial submissions.

CJ Follini's concept of the investor or financial professional who understands the full spectrum of financial instruments, asset classes, tax strategies, and planning tools — rather than deferring entirely to hyper-specialists for each domain. The financial generalist knows enough about real estate, venture capital, estate planning, retirement accounts, tax optimization, and debt management to ask the right questions, identify the right specialists, and make integrated decisions. CJ argues this is the model the AI era rewards: LLMs and agentic AI are themselves generalist technologies, and the human who can orchestrate specialist agents across financial domains outperforms the human who outsources each domain to a separate professional with misaligned incentives.

A large language model that has been further trained on a domain-specific dataset — in Videra's case, behavioral health conversations between providers and patients — to make its outputs more accurate and contextually appropriate for that domain than a general-purpose model.

Pays for your own losses: business interruption, data restoration, model retraining, runaway infrastructure costs.

If you cannot describe your product in five words or less in H1 text on a website, it is a bad product. Josh Furstoss's test for whether a product is sufficiently specific. One core feature. Everything beyond that dilutes the offer, expands the surface area to defend, and signals that the founder has not yet identified the actual problem they are solving.

Emberos's term for the specific, strategic recommendations it generates to correct AI misrepresentation — ranging from FAQ creation to PR placements to YouTube video production.

A clinical presentation in which a person shows little or no emotional expression — reduced facial movement, monotone voice, minimal gesture. An early visual indicator of several neurological and psychiatric conditions including Parkinson's disease and severe depression.

A lead generation model where all vendors in a category pay the same fixed price per lead, regardless of company size or budget. Eliminates the bidding dynamics that bias results toward whoever spends most — preserving matchmaking integrity and ensuring recommendations are based on fit, not payment.

Sensory deprivation tank. Dan credits it with "tens of millions of dollars" worth of ideas.

The mental state of full engagement when you know the next step, and the next, and momentum is compounding. Characterized by energy, direction, and the feeling that today you will finish this thing. When broken by an unresolved technical blocker, vibe coding projects don't pause — they die. The project sits on the laptop until the founder's attention moves on. Protecting flow state is the design principle behind On Demand Human: a 15-minute expert intervention at the right moment is worth more than any platform feature.

Fast-moving consumer goods. The packaged goods sector where Grobler spent two decades.

Follow links pass SEO authority from the linking domain to the destination. No-follow links signal to crawlers that authority should not be transferred — used on sponsored content and untrusted sources.

The constant background preoccupation with food and eating that many people with appetite dysregulation experience. GLP-1 medications are highly effective at reducing or eliminating it.

An enterprise go-to-market model in which engineers are embedded directly with the customer post-sale to accelerate implementation, drive adoption, and establish time-to-value. Knapsack adopted this model after discovering that enterprise clients were stalling 4–6 months after signing — not because the product didn't work but because implementation required internal IT resources the customer hadn't budgeted. Forward deployment removes that friction by making the vendor's engineers the implementation team.

Dobbin's term for the structured 15-dimension description of a company — culture, values, brand, strategy, and objectives — that powers all of its judgment.

The process of taking a pre-trained large language model (such as GPT-4, Claude, or Gemini) and continuing its training on a smaller, domain-specific dataset to specialize its behavior for a particular use case without training a model from scratch. Fine-tuning allows companies to capture the reasoning and language capabilities of a foundation model while customizing it for specific vocabularies, response formats, constraints, or domain knowledge. Loman.ai fine-tunes foundation models for the specific context of restaurant phone conversations - teaching them the nuances of taking food orders, handling 86'd items, managing reservation requests, and maintaining the appropriate tone for different restaurant types. Christian's approach: test every major foundation model, identify which performs best for each specific sub-task within the product, and fine-tune for production performance.

Areas of the startup journey where a founder does not know what they do not know — the most common being what validation actually requires, what investors need to see before taking a meeting, and how to distinguish genuine market signal from polite social feedback.

The ability to see patterns in your own thinking, behavior, stress, and decisions rather than simply reacting in the moment.

A go-to-market strategy in which the founder is the primary public voice — doing podcasts, long-form content, and community building rather than delegating outbound to a sales team. Especially effective when domain credibility and authentic expertise are the primary differentiators.

When founders publicly build credibility through content and personal brand.

A senior technical leader hired part time, giving startups CTO-level engineering leadership while keeping equity and payroll cost down. One of Cause of a Kind's core service offerings.

Scattered photos, chats, and media stored across platforms without structure.

Clyde's pricing ladder, upgrading users through product usage.

An analysis that determines whether a product, service, or process can be commercialized without infringing on valid patent rights held by third parties. An FTO clearance does not mean the inventor's idea is patentable — it means the inventor can build and sell without risk of infringement litigation. Ophir frames FTO as the first question every founder should answer before raising money, hiring engineers, or talking to customers: if someone else owns the rights to your core mechanism, finding out on day zero costs almost nothing; finding out at Series A or after a lawsuit is potentially company-ending.

A pricing model where a product is offered free at a basic tier with paid upgrades. Common for top-of-funnel acquisition but a leading cause of high churn in crowded AI tool markets.

The patient-facing operational layer of a practice: phones, scheduling, eligibility, referrals, prior authorizations, intake.

The handful of companies training the most capable models, including Anthropic, OpenAI, and Google.

A hair transplant method where individual follicles are extracted and transplanted one by one. Less invasive than strip surgery and the dominant technique in the medical tourism market.

Kindling Solutions's full build: a custom AI team installed across the business so output scales without the founder in the middle.

The full financial statements behind the headlines: income statement, balance sheet and cash flow statement.

The quality and conversion rate of users moving through awareness, activation, and retention. A key metric founders should prove before returning to investors — demonstrate the funnel works, then raise from traction rather than hypothesis.

Gainable’s engine that reads data sources and generates apps.

A setup where two neural networks compete: a generator creates fake data and a discriminator tries to spot it, which pushes the generator to get more realistic.

Recommended care a patient hasn't received: overdue screenings, missed well visits, skipped vaccinations. Closing them is directly tied to reimbursement in US value-based arrangements.

A German compound word (Geist = spirit/mind, Blitz = flash/lightning) describing a sudden flash of insight or inspiration - an aha moment where a fully formed idea, solution, or memory surfaces unexpectedly. Related to but distinct from the common eureka moment: Geist Blitz can also describe the experience of a thought or story that disappears mid-sentence and then resurfaces later in full detail, seemingly having incubated in the subconscious. Ryan and Nicolas discuss this phenomenon at the opening of the episode, arriving at the word after Ryan describes the experience of a story evaporating mid-interview and then returning intact minutes later. Nicolas' aunt, a German teacher, is implicitly invoked as the authoritative source on correct usage.

AI that creates new content, such as text, images, music, code, or video, rather than only analyzing existing data.

AI-generated music personalized for specific emotional states, relationships, or identity contexts. In Ziah Orion's work at Deep Gem Interactive, generative music is not content — it is a personalized emotional artifact created for a specific person and relationship (grief, love, identity), with the specificity itself being the source of value.

Hot, mineral-rich water deep underground; Cornwall's brines carry dissolved lithium.

A type of diabetes that develops during pregnancy in patients who did not previously have the condition, caused by hormonal changes that impair insulin function. Gestational diabetes increases the risk of complications including large-for-gestational-age infants, premature birth, and cesarean delivery, and elevates the mother's long-term risk of developing type 2 diabetes. Management requires regular blood glucose monitoring, dietary adjustment, and in some cases medication. Remote patient monitoring platforms like Babyscripts extend glucose tracking into the home, enabling continuous surveillance of blood sugar levels rather than relying solely on clinic-based testing at scheduled appointments.

The gi is the uniform, a roll is a sparring round, a sweep reverses position from the bottom, a submission ends the round, closed guard and side control are positions, and purple belt is the intermediate rank both hosts hold.

A Git feature that allows multiple working copies of the same repository to exist simultaneously on the filesystem, each checked out to a different branch. In AI-native development, worktrees enable parallel agent execution: separate agents work on separate requirements within the same codebase simultaneously, each on its own branch, without interfering with each other. Tyler describes worktrees as 'magic' for velocity — you're not waiting for one task to complete before starting the next. When all branches pass testing, they merge sequentially. Worktrees transform sequential agent workflows into parallel pipelines.

A platform for storing and versioning code.

A Bloom pro feature that connects a Bloom project to a GitHub repository. Code changes made in Bloom push to the repo; changes made in external tools like Cursor or Claude Code pull back into Bloom.

How a specific meal moves blood sugar, viewable in the Qalzy app using data from Apple Health or Google Health Connect.

A newer role applying AI and automation to traditional go-to-market work. Same job description as a GTM pro, plus the tooling.

A bloated code file with too much logic that gums up a codebase.

An all-in-one marketing and CRM platform used here for publishing.

The shortest route a user takes through your product to the moment of value and revenue.

Deliberate design resistance that forces users to have skin in the game, improving output quality and creating accountability signals.

The pattern of continuing to invest in a strategy that has already failed, hoping additional capital will change the outcome. A founder trap where sunk cost psychology overrides clear evidence that the go-to-market hypothesis is wrong.

Google's free tool for submitting URLs for manual indexing and tracking organic search performance. The primary workaround for low-DA sites waiting for Google's crawler to discover new content.

AI agents running under defined rules, permissions, and data protections.

A family of large language models built on the transformer architecture and trained to generate human-like text from a prompt.

OpenAI's image model (branded Images 2.0), used in this workflow for likeness-preserving anchor frames that give the video model fixed visual endpoints to bridge between.

A thin AI product built on top of a foundational LLM without proprietary data, unique distribution, or genuine differentiation. The category of AI product most founders are building and fewest investors want to fund as foundational models grow more capable. Lacks the defensible IP or distribution moat required for long-term pricing power.

A chip built to run many calculations in parallel. Originally made for graphics, it is now the workhorse for training and running AI.