Why we are turning Founders DNA into GitHub for founders

TL;DR

The product as it stood, six months ago

When we shipped the Founders DNA alpha, it did one thing well. A founder registered a virtual business — a cloud kitchen in Mumbai, a SaaS for dentists, a print-on-demand brand — and the system delivered ninety days of business events to them on Telegram. Supplier shocks. Cash crunches. Hiring crises. Regulatory surprises. Each event came with three choices and a stopwatch. The founder picked one, the local LLM scored the decision across six dimensions, and over thirty or forty decisions a Founder DNA radar emerged.

The dimensions were the right ones — Risk Calibration, Capital Discipline, Growth Instinct, Operational Rigor, People & Leadership, Crisis Response. The simulator was the right idea. The privacy posture (local LLM, no third-party API for decision data) was the right posture. The Telegram delivery was the right surface.

It was an honest test of one specific thing: how a founder thinks under pressure.

The thing it did not measure — and could not — was what the founder actually did the rest of the day.

The conversation that changed it

I spent eleven years at Microsoft, the last several of them building Copilot. When I left to start CopilotVerse and ship Founders DNA, the people I called first were the investors and accelerator partners I had got to know in Bangalore, Mumbai, and Singapore. The pitch was simple: a verifiable behavioural profile of a founder, before you fund them.

The reaction was uniformly enthusiastic about the simulator and uniformly frustrated about its limit. The line that kept landing was some version of this: "I love what the radar tells me about how they decide. The thing I cannot get from any source today is what they actually did this week."

Decks lie. Resumes lie. A founder's verbal answer to "tell me about a time you handled a crisis" lies — not maliciously, but selectively. Even a really good simulator can be approached as a test, and tests can be gamed by people who treat them as tests. The thing that is hardest to fake is a record of action that builds itself in the background of someone's day.

That is what GitHub is, for code.

The insight

GitHub did not set out to be a credibility platform. It solved versioning and collaboration for software projects, and the credential — the green-square contribution graph that every senior engineer now treats as a kind of resume — emerged because every push was an authentic artifact of work. Nobody had to be persuaded to commit. The product they used daily produced the profile as a byproduct.

That ordering matters. If GitHub had launched as "Build your developer credential," nobody would have come back daily. The credential exists because the daily utility came first.

To build the same thing for founders, you cannot lead with "build your founder credential." You have to lead with a daily utility that founders genuinely want — and let the verifiable profile emerge underneath.

The daily utility we picked is action coach plus accountability. The data backbone we picked is broader than a chat: it is a universal business event capture system.

The new shape: the founder commit

Most authentic execution data already lives in the tools founders use every day. GitHub. Linear. Notion. Stripe. Razorpay. Slack. LinkedIn. Calendly. Vercel. Mailchimp. Shopify. Tally. WhatsApp Business. Plus the things that do not have a clean digital trail — a hand-painted sign, a vendor handshake, a voice memo from a customer call.

The new architecture has one universal table — founder_events — and three sources that all write into it:

• OAuth integrations (the primary commit source)
• Manual proofs sent to the Telegram bot (text, photo, voice note, link)
• Sim decisions (the existing Judgment DNA track)

Each row carries a normalized event type, a timestamp, a payload, and an AI-generated summary. The contribution heatmap is computed from this table. The activity feed reads it. The scoring engine consumes it. The whole UI is polymorphic over a single typed stream — which means we can ship many integrations behind the same scaffold without rewriting anything downstream.

The first wave of native connectors going live with the beta:

#	Integration	Captured events
1	GitHub	commits, releases, PRs merged, repos created
2	LinkedIn	public posts
3	Notion	pages created, page updates
4	Google Forms	form responses received
5	Calendly	meetings booked, meetings completed
6	Stripe	charges, subscriptions, refunds, customer.created
7	Razorpay	payments, payouts
8	Vercel	deployments

The second wave, rolling through Phase 1: Google Calendar, Gmail, X/Twitter, Slack, Discord, Linear, GitLab, Netlify, Railway, Cal.com, Typeform, Tally Forms, Mailchimp, ConvertKit, Resend, Substack, Plausible, PostHog, Google Analytics, HubSpot, Pipedrive, Airtable, Paddle, Lemonsqueezy, Google Docs.

Plus a generic Zapier / Make / n8n bridge from day one — any tool that does not yet have a native connector can wire in through a webhook URL we generate per user. We are not gating "universal" on building twenty-three integrations before launch.

Two radars, one founder

The most important consequence of this design is that we now have two parallel measurements of the same person. The simulator scores judgment under pressure. The activity graph scores execution.

Judgment DNA is the existing six-dimension radar — Risk Calibration, Capital Discipline, Growth Instinct, Operational Rigor, People & Leadership, Crisis Response. Sourced from the simulator. Unchanged.

Execution Profile is a new four-dimension radar — Velocity, Consistency, Depth, Reach. Sourced from founder_events. Streak math is local; the LLM judges depth (was the artifact substantive?) and reach (did it generate a response from someone with no reason to be polite?).

The two radars are different muscles, and that is the point. A founder can score 92 on Crisis Response in the simulator and have a flat heatmap the rest of the month. A founder with green squares every day for ninety days might still freeze when the simulator throws a supply-chain shock at them. One number hides the contradiction. Two radars make the divergence visible.

I think this is the single most important design decision in the new product. Investors do not just want to know who decides well or who executes well — they want to know whether the same person does both. We separate them deliberately, on the same page, side by side.

I wrote a whole post on the reasoning behind this — why one DNA score is not enough.

GitHub-shaped, deliberately

We are going all-in on GitHub's open-source design system, @primer/react, plus @primer/octicons-react for icons. The vocabulary is the experience.

GitHub	Founders DNA
Repository	Venture (one of many per founder)
Commit	Founder event — webhook, manual proof, or sim decision
README	Venture About page (idea, stage, public links, current state)
Issues	Open tasks + sim scenarios awaiting decision
Pull Requests	Decisions made (a closed PR = a decided choice)
Actions	Scheduled workflows — daily task fire, integration syncs
Stars / followers	Mentors, peers, watching investors
Insights	Judgment DNA + Execution Profile + integration-derived charts
Profile	Anon founder page with pinned ventures, dual radars, contributions heatmap
Marketplace	Integrations marketplace

The page tree mirrors GitHub's exactly: /founder/<slug>, /founder/<slug>/v/<venture>, /founder/<slug>/v/<venture>/issues, /founder/<slug>/v/<venture>/pulls, /founder/<slug>/v/<venture>/insights. Every route a founder uses inherits the muscle memory of a tool they already know.

We could have invented our own design language. We chose not to, because the product is the analogy. Reverse-engineering GitHub's visual feel with a custom system would have made the analogy claimed; using Primer makes it felt.

Why HS and college as the wedge

The first version of Founders DNA was pitched at "founders" generically — anyone with a venture worth simulating. That was a mistake. The wedge is wrong if your first user has nothing to come back for.

We are starting with high-school and early-college aspirers running mostly-virtual ventures. Side projects, school clubs, content brands, course experiments, indie SaaS, food carts, tutoring businesses. The reasons stack:

• Highest density of "want to build, nothing to lose" — they are not optimising for next quarter's pitch; they are figuring out whether they want to do this at all.
• Cleanest fit for the contributions metaphor — virtual ventures throw off a lot of digital signal (LinkedIn posts, Notion pages, Google Forms, GitHub commits) and very little PII.
• Most authentic reason to use a daily action coach — "I am building toward something" is a more durable loop than "I am validating a credential."
• Network effect upstream — the founders who come of age inside Founders DNA show up to investors and accelerators in a few years already carrying the credential. Reverse-engineering that arrival is harder than building it forward.

Investors and accelerators get a great product the moment a critical mass of profiles exists. They do not need to be the first user.

I wrote a separate post on why the wedge is high-school and college — including the wedge-vs-positioning distinction that took us a while to internalise.

Privacy, by default

A founder profile that aggregates Stripe charges, GitHub commits, LinkedIn posts, and Calendly attendees is a privacy minefield if you do it wrong. We default to four properties:

• Anon-public. Your public profile is founder-4823, not your name. The heatmap and the radars are visible; the identity is not.
• Identity unlocked selectively. You generate a magic link granting a specific viewer (an investor, a mentor, a future co-founder) full identity plus venture access for N days. They see who you are; the world does not.
• Local LLM. The simulator and the scoring engine run on Llama-class models on your own infrastructure. Your decisions, your reflections, your business state never leave to a third-party API.
• Aggressive payload redaction. Webhook payloads carry PII — Stripe customer emails, Calendly attendees, Slack message bodies. We store the founder action and aggregate counts; we drop the rest at the integration layer before it ever lands in the database.

This is not a side feature. The product cannot exist without it.

What is live today, vs Q3 2026

I want to be precise about which parts of the new product are shipped and which are coming.

Live today (private alpha):

• The simulator — 90-day plan engine, 7 industry verticals, LLM event generation, fallback to template events when Ollama is unreachable
• 6-dimension Judgment DNA scoring, with a 5-layer evaluation framework that validates coherence, category accuracy, scoring consistency, financial realism, and milestone logic
• Telegram bot delivery (events out, decisions captured)
• Fast-forward simulation (compress 90 days into minutes)
• Downloadable DNA Card (the existing single-radar credential)

Building for Q3 2026 beta:

• Integrations marketplace and the eight first-wave OAuth connectors
• Generic Zapier / Make / n8n bridge
• The new founder_events table and the contributions heatmap
• Execution Profile (Velocity, Consistency, Depth, Reach)
• Anon founder profile pages, multi-venture portfolio
• The Primer-based UI shell — repo / issues / pulls / insights tabs
• LLM-driven cadence engine (decides daily what fires for each founder: task, sim event, both, or rest)
• Manual proofs on Telegram (text, photo, link; voice in Phase 2)
• Identity-unlock magic links (Phase 2)

If you are reading this and you are inside the alpha, nothing about your existing simulator data goes away — your DNA radar carries forward unchanged into the new product as the Judgment track.

The ask

Three buckets, in priority order:

• Design partners. If you are an aspiring founder — high-school, college, career-switching, building something on the side — we will give you founding-member status and a weekly call with me during the first eight weeks. In exchange we ask for written feedback every week and permission to use anonymised metrics in our public results.
• Seed investors. We are raising to scale infrastructure (the integration scaffold takes serious engineering), build the Q3 beta team, and run two paid pilots — one with an accelerator, one with a business school replacing capstone projects.
• Accelerators and business schools. If you screen 500 applicants for 20 spots, or grade 100 capstone projects every term, the dual-radar profile is the artifact you have always needed. Two pilot slots in the Q3 beta cohort.

Reserve a profile or get in touch →

A note on what stays and what changes

The mission is the same one we started with: a verifiable behavioural profile of a founder, before you fund them. What changed is the realisation that "behavioural" has to include real-world action, not just simulated decisions — and that the only way to capture real-world action authentically is to be the daily tool that founders use, with the credential as a byproduct.

GitHub got there for code. We are doing it for founders.

— Karthik