AI Cofounders | Blog

AI cofounder vs hiring a technical cofounder: cost, equity, and what you actually get

Fri, 12 Jun 2026 00:00:00 GMT

“How do I find a technical cofounder?” is one of the most-asked questions in every founder community — and in 2026 it has a genuinely new answer. Before you spend a year searching and 50% of your cap table paying, run the numbers on both options.

The real cost of a technical cofounder

Equity: the standard ask is 30–50%. On a company that exits for even $1M, that’s $300k–$500k — the most expensive hire you will ever make, agreed to at the moment you know the least.

Time: founder-dating platforms, meetups, and cold DMs typically take 6–12 months to produce a committed partner. That’s a year of not building.

Risk: Harvard research (Noam Wasserman, The Founder’s Dilemmas) attributes roughly 65% of startup failures to founder conflict. A cofounder you found on a matching platform three months ago is a marriage to a stranger.

And the quiet truth: most “I need a technical cofounder” searches are really “I need software built and I can’t afford an agency.” That’s an execution problem, not a partnership problem.

What you actually need, by stage

Validating (pre-build): you don’t need a CTO to find out if anyone wants this. You need market research, a landing page, and a waitlist. AI does all three today — start with a free idea teardown (no signup).

MVP: AI code generation in 2026 comfortably produces working web apps — scaffolded, deployed, iterated. An AI tech cofounder plans the 3-day MVP sprint, scaffolds the app, and pushes to your GitHub. You own the repo and 100% of the company.

Scale (real users, real load, real security): this is where senior human engineering judgment genuinely matters. Recruit it — as a CTO hire or late cofounder — from traction. “I have 500 paying users and need help scaling” attracts world-class engineers. “I have an idea and a deck” attracts nobody.

The comparison, honestly

	Technical cofounder	AI cofounder
Cost	30–50% equity, forever	~$29–149/month, cancel anytime
Available	After 6–12 months of searching	This afternoon
Skills	One person’s stack and opinions	Product + tech + marketing + sales + ops + finance
Commitment	Total (the irreplaceable part)	None — it’s software
Conflict risk	The #1 startup killer	Zero (it can’t quit, sulk, or fork the repo)
Investor signal	Strong	Neutral — but traction beats team slides
2am conviction	Yes	No — conviction stays your job

The right column doesn’t dominate the left. A great human technical cofounder — one you’ve worked with, who shares your vision — is still the strongest setup in startups. But a mediocre technical cofounder acquired out of desperation is worse than none, and you can’t un-give equity.

The sequence that wins in 2026

Validate with AI — teardown, market research, landing page, waitlist. Cost: ~$0.
Build the MVP with an AI cofounder team — code scaffolded and pushed, marketing and outreach running in parallel, every action approved by you. Cost: a SaaS subscription.
Get revenue.
Then decide — many founders discover at step 3 that they never needed to give away half the company. Those who still want a human CTO now recruit from strength, and can offer 5–15% instead of 50%.

The technical cofounder search used to be a gate: no engineer, no startup. AI execution removed the gate. What’s left is the part that was always the real test — whether you can find a problem worth solving and stay with it.

That part is still yours. The rest, your AI cofounders can start on today.

AI cofounder vs human cofounder: which one do you actually need?

Fri, 12 Jun 2026 00:00:00 GMT

I build AI cofounders for a living, so you’d expect me to tell you an AI cofounder beats a human cofounder every time. It doesn’t. They solve different problems, and picking wrong costs you either 50% of your company or months of stalled execution.

Here’s the honest version.

What an AI cofounder actually is

An AI cofounder is an AI system that does cofounder-level work — not cofounder-level commitment. The good ones go beyond chat: they research your market, write and send your outreach, build and deploy your landing pages, model your finances, and remember the context of your business across months of work.

The term got popular in 2024–2025 as solo founders realized that general chatbots weren’t enough. A chatbot answers questions. An AI cofounder owns a function — product, marketing, sales, tech, operations, or finance — and produces the deliverables that function is responsible for.

What an AI cofounder is not:

Not a legal partner. It holds no equity, signs nothing, and carries no fiduciary duty.
Not a believer. It won’t take a pay cut for two years because it believes in you.
Not your network. It can draft the investor email; it can’t be the warm intro.

What a human cofounder gives you that AI can’t

Let’s start with the side that doesn’t favor my product.

Skin in the game. A human cofounder with 30–50% equity is financially destroyed if the startup fails. That alignment changes behavior in ways no software can replicate — they’ll take the 2am support call, front their own money, and push through the month you want to quit.

A counterweight with veto power. An AI will challenge your assumptions if it’s built to (ours is), but it can’t stop you. A human cofounder can look you in the eye and say “we are not pivoting again” — and make it stick.

Credibility with investors. Many VCs still treat a solo founder as a risk flag. A strong technical cofounder on the cap table de-risks the round in a way an AI subscription doesn’t.

Network and luck surface. Cofounders bring their former colleagues, their Twitter following, their old customers. That’s distribution you can’t subscribe to.

If you have access to a great human cofounder — someone you’ve worked with before, with complementary skills, who wants the same company you do — take them seriously. That’s still the strongest configuration in startups.

What an AI cofounder gives you that a human can’t

You keep 100% of your equity. The median cofounder split is 50/50. An AI cofounder team costs less per month than a single dinner-and-drinks recruiting pitch, and it never vests.

No search, no breakup risk. Finding a cofounder takes 6–12 months on average, and cofounder conflict is one of the top reasons startups die (Noam Wasserman’s research at Harvard put founder conflict behind roughly 65% of startup failures). An AI cofounder is working within the hour and can’t rage-quit with half your codebase.

Six functions instead of one. A human cofounder covers one, maybe two domains. An AI cofounder team covers product, tech, marketing, sales, operations, and finance simultaneously — with each one applying real frameworks (RICE, SPIN Selling, OKRs, Bessemer SaaS metrics) instead of vibes.

Volume of execution. This is the one founders underestimate. A human cofounder writes one landing page this week. An AI cofounder team drafts the landing page, the 7-touch outreach sequence, the 30-day content calendar, and the 12-month cash flow model — this afternoon — and you spend your time approving and steering instead of producing.

The honest decision matrix

Your situation	What I’d pick
You’ve found a great human cofounder you’ve worked with before	Take the human. Use AI to multiply both of you.
You’re searching for a cofounder because “you’re supposed to have one”	AI cofounder. A mediocre human cofounder is worse than none.
You’re non-technical and need production software at scale	Eventually a human CTO — but validate with AI first so you recruit from strength.
You’re a builder who hates marketing/sales	AI cofounder team now; hire humans when revenue justifies it.
You’re pre-idea, exploring	AI. Don’t give away equity before you know what the company is.
You’re raising VC and investors want a team	Recruit the human — and walk in with the traction your AI team helped you build.

The hybrid that actually wins

The framing “AI vs human” is slightly wrong, the way “calculator vs accountant” was wrong. The configuration winning right now in 2026 is the solo founder + AI cofounder team: a single human with full ownership and conviction, multiplied by AI that executes across every function — with the human approving every action.

You can always add a human cofounder later, from a position of strength: working product, real users, real revenue. You can’t easily subtract one.

FAQ

Can an AI cofounder really replace a human cofounder? For execution — research, marketing assets, outreach, code scaffolding, financial models — largely yes. For equity-level commitment, investor signaling, and network, no. Most solo founders need the execution far more urgently.

Do investors take solo founders with AI teams seriously? More every quarter. Traction beats team composition: a solo founder with revenue outranks a complete founding team with a deck. AI execution is how solo founders get to that traction.

How much does an AI cofounder cost vs a human one? A human cofounder typically costs 30–50% equity. AI cofounder tools run $20–$200/month. If your company ends up worth anything at all, the equity was the most expensive thing you ever spent.

What’s the catch with AI cofounders? Judgment is still yours. An AI team multiplies your direction — including a bad one. That’s why ours requires founder approval on every action: the AI proposes, you decide. If you want to see how that feels, run a free teardown of your idea — no signup, takes a few minutes.

The best AI cofounder tools in 2026 — compared by someone who builds one

Fri, 12 Jun 2026 00:00:00 GMT

Full disclosure up front: I build aicofounders.co, so I’m a competitor in this list. I’ll keep every claim about the other tools sourced from their own public pages, and I’ll tell you who each tool is genuinely right for — including when it’s not mine.

The “AI cofounder” category exploded between 2024 and 2026, and the names are confusingly similar: aicofounder.com, cofounder.ai, cofounder.co, aicofounders.co. They are four different products with four different theses. Here’s the map.

TL;DR table

Tool	What it really is	Stage	Best for
aicofounder.com	Guided research & planning with one AI	Live, 80k+ founders	Figuring out what to build
CoFounder.AI	Six AI specialists over iMessage/WhatsApp	Waitlist	Phone-first founders willing to wait
Cofounder.co	Agent orchestration platform for running a company	Live	Technical founders who want to wire their own agents
ChatGPT / Claude	General-purpose chatbot	Live	Everything and nothing in particular
aicofounders.co	Six AI cofounders that execute, with founder approval	Live (closed beta)	Solo founders who need the work done

aicofounder.com — the research guide

Formerly Buildpad, rebranded in February 2026. It walks you through structured phases: brainstorm, validate, research, plan — with market research that cites its sources and a visual canvas for your product. Their 80,000+ founder user base is real social proof, and the guided flow is genuinely good at stopping you from building something nobody wants.

The limit: it’s one AI focused on research and planning. When the plan is done, the landing page, the outreach, the code, and the financial model are still your job.

Pick it if: your main risk is building the wrong thing, and you’re happy to execute everything yourself.

CoFounder.AI — the messaging-first team

CoFounder.AI (“The AI CoFounder — zero equity, all execution”) puts six AI specialists — growth, product, sales, finance, marketing, operations — on iMessage and WhatsApp. The thesis is the same one I bet on: founders need a team, not a chatbot.

The limit: as of June 2026 it’s a waitlist, with execution starting after onboarding. Frameworks, progress tracking, and the founder-control model aren’t publicly documented yet.

Pick it if: you want your AI team in your texting app and don’t mind waiting for access. (Here’s my detailed side-by-side with CoFounder.AI.)

Cofounder.co — the agent orchestration platform

Cofounder.co is the most ambitious framing: “run an entire company with agents” — engineering, sales, marketing, design, finance, ops, with infrastructure, analytics, and Stripe payments wired in.

The limit: it’s a platform, not a team. You get a runtime and building blocks; the workflow design is on you. Power and responsibility scale together.

Pick it if: you’re technical, you enjoy designing agent workflows, and you want maximum control over the machinery.

ChatGPT / Claude — the generalist

The tools everyone already has. Brilliant for one-off questions, drafts, and thinking out loud.

The limit: no persistent business memory across functions, no deliverables shipped to real tools, no methodology, and famously agreeable — your hat-for-ducks idea is always “a great niche!” A chatbot answers; it doesn’t own anything.

Pick it if: you want a thinking partner and you’ll do all the structuring, remembering, and executing yourself. (Longer version: AI Cofounders vs ChatGPT.)

aicofounders.co — the team that executes (mine)

My bet is different from all of the above: a solo founder doesn’t need more advice or more infrastructure — they need a team that does the work, under their control. So aicofounders.co gives you six AI cofounders (Product, Tech, Marketing, Sales, Operations, Finance) that:

validate ideas with live research on Reddit, Hacker News, Product Hunt, and Google Trends
deploy real landing pages, draft outreach sequences, scaffold code, build financial models
track everything on 12 live dashboards (Business Model Canvas, Sprint Board, Sales Pipeline, Cash Flow, KPIs)
work with 15+ named frameworks — Lean Startup, RICE, SPIN Selling, BANT, OKRs, Bessemer SaaS metrics
ship through 20+ integrations (Gmail, GitHub, Mailchimp, HubSpot, Stripe…)

— and the part I refuse to compromise on: every action is a proposal you approve or reject before it runs. The AI executes; you stay the founder.

The limit: it’s a closed beta, so you request access rather than swipe a card. And if all you want is research, aicofounder.com’s guided flow is more polished for that single job.

Pick it if: you’re a solo founder whose bottleneck is execution — the marketing that doesn’t get done, the outreach you keep postponing, the model you never build.

How to choose in 30 seconds

“I don’t know what to build” → aicofounder.com, or start with a free AI teardown of your idea (no signup).
“I know what to build, I can’t do it all alone” → aicofounders.co.
“I want to engineer my own agent company” → cofounder.co.
“I just want to chat through ideas” → ChatGPT or Claude.

Whatever you pick: the founders winning in 2026 aren’t the ones with the best AI subscriptions. They’re the ones who turned AI output into shipped artifacts — pages live, emails sent, code pushed. Choose the tool that gets you to shipped.

How to validate a startup idea with AI — free, in about 30 minutes

Fri, 12 Jun 2026 00:00:00 GMT

Most founders validate their startup idea by asking ChatGPT “is this a good idea?” and hearing “what a great niche!” That’s not validation — that’s a compliment machine.

Real validation answers four questions with evidence:

Does the pain exist? (Are real people complaining about this, in public, recently?)
Who exactly has it? (A reachable tribe, not “everyone who…”)
What do they do about it today? (Competitors and workarounds — both are good news)
Will they pay? (Is money already moving in this space?)

Here’s how to get evidence-based answers using AI, for free, in about 30 minutes.

Step 1: Hunt the complaint, not the compliment (10 min)

Go where your audience already complains: Reddit, Hacker News, niche forums. The AI move is to use a model with web search and force it to cite:

“Search Reddit and Hacker News for people describing this problem: [your problem]. Give me direct quotes with links, dated within the last 12 months. If you can’t find at least 5, say so.”

The last sentence is the important one. You’re trying to make “there’s no demand” a possible answer. If the AI can’t find recent, specific complaints, that’s your result — cheaper to learn now than after three months of building.

Step 2: Name the tribe (5 min)

“Busy professionals” is not a tribe. “Solo therapists who hate writing post-session notes” is. Push the AI:

“Based on those complaints, describe the single most specific group with this pain. Where do they hang out online? What words do they use for the problem?”

The words matter — they become your landing page headline and your search keywords.

Step 3: Map competitors and workarounds (10 min)

“List products that solve this today, with pricing. Then list the manual workarounds people describe (spreadsheets, VAs, duct tape). What do users complain about in each?”

Two traps here:

“No competitors” is usually a red flag, not an opportunity. It often means no budget exists.
The workaround is your real competitor. If people solve it with a free spreadsheet, your $49/month tool fights the spreadsheet, not the other SaaS.

Step 4: Force a verdict (5 min)

This is the step everyone skips, because chatbots are agreeable by default. Force it:

“You are a skeptical product advisor who has seen 1,000 failed startups. Given the evidence above, give me: a GO / NO-GO / PIVOT verdict, the 3 biggest risks, and the cheapest possible test for the riskiest assumption. Do not soften the verdict.”

You’re not asking permission to build. You’re asking what would have to be true — and what the cheapest way to check it is.

The traps that invalidate your “validation”

Leading the witness. Ask “what problems do you have with X?” — never “would you use a tool that does Y?”
Validating the solution instead of the pain. People lie about what they’d use; they don’t lie about what already hurts.
Counting upvotes as demand. Likes on “I’d love this!” are not pre-orders. Money, emails, and waitlist signups are.
One-and-done. Validation isn’t a gate you pass once; the verdict updates with every new piece of evidence.

Or run the whole thing in one shot (free)

I turned this exact process into a free tool: the AI startup idea teardown. You paste your idea, and the Product cofounder from aicofounders.co runs the full diagnostic — honest verdict, pain level, the specific tribe, named competitors, real risks, and 5 concrete actions for this week.

No signup, takes a few minutes, and the verdict is deliberately blunt — it will tell you NO-GO when the evidence says NO-GO. You can also browse public teardowns other founders have run to calibrate what honest validation looks like.

Worst case, you lose 5 minutes. Best case, you avoid losing 3 months.

86 board items, 0 shipped artifacts — the diagnostic that rewired my product

Wed, 03 Jun 2026 00:00:00 GMT

The data that stopped me

Two months into the closed beta of aicofounders.co. Two active users. I pulled the numbers on what they’d actually done inside the product.

Vincenzo: 84 messages with the AI cofounders. 44 strategic board items populated. 0 real-world artifacts shipped.

Valerio: 45 messages. 42 board items. 0 artifacts shipped.

Between them: 130 messages, 86 board items, zero things produced in reality. No landing pages deployed. No outreach emails sent. No GitHub pushes. No social posts made.

The product technically worked. The tools were wired. They could have executed — they just didn’t.

I sat with this for three days. The instinct was: nobody wants this product, kill it. But the engagement was high. The output volume was high. The piece that was zero was crossing the line from thinking to shipping.

That’s not a “users don’t want this” diagnostic. That’s a UX problem: the product makes thinking easy and shipping invisible.

Reading the data carefully

Here’s what made it interesting. Both users worked through their first session for 90+ minutes. They populated multiple boards (Idea Teardown, Validation, Sprint, Pipeline). They generated cold email drafts, landing page copy, sprint plans, financial models.

All of those outputs are executable — the tool to ship each one exists in the product. There’s a “deploy landing page” tool, a “send via Gmail” tool, a “push to GitHub” tool. They just sit in a Tasks panel users discover about 30% of the time.

When the Sales AI drafted a cold email and said “want to send this?”, the user had to mentally rebuild the context: which tab is the tasks panel in, what does approval look like, do I need to connect Gmail first. Three layers of friction between draft and ship.

That friction is the difference between “AI that helps me think” and “AI that helps me ship.” Different products. Different value props.

What I shipped to close the gap

30 days of work, no new capabilities, only UX:

Action buttons on every deliverable card. When the Sales cofounder produces an email draft, the card now has a primary “Send via Gmail” button right under it. Not in a Tasks panel. Inline. Same for “Push to GitHub”, “Post to LinkedIn”, “Deploy live”.
A system rule I called EXECUTE-OR-OFFER. Every cofounder’s system prompt now has a forced rule: when you produce an executable artifact, you MUST end your response with a specific ship-offer. Not “let me know what you think” — “want me to send this to david@acme.co right now?”
A “Shipped” panel. New side panel listing every real-world artifact the cofounders have actually produced. Sent emails, deployed pages, pushed code, posted social. Empty until you ship something. Becomes a visible scoreboard.
A reframed onboarding question. “What does it do?” became “What do you want SHIPPED this week?” The first cofounder turn aims at the founder’s stated weekly outcome, not 12-month strategy.
Connection-aware buttons. If Gmail isn’t connected, the button reads “Connect Gmail → Send” with one-click OAuth.
Preview modal for high-risk actions. Click “Send via Gmail” → modal opens with editable subject, recipient, body. Cancel / “Looks good — send.”
Background nudge cron behind a feature flag. Daily cron that surfaces what’s overdue. Gated by an env var — defaults off.

Total: 17 files modified, ~600 lines added, single branch.

The harder lesson

Shipping “no new features” is uncomfortable. It feels like inactivity. The marketing department of your own brain says “but I need something new to post about.”

But the diagnostic was clear. The capabilities existed. The hierarchy was wrong. If your data shows users engaging but not converting to action, your next sprint probably isn’t more features. It’s surface area redesign.

The hardest features to ship are the ones that change user behavior, not the ones that add surface area.

What I’m watching now

Did the change move the needle? The honest answer is: I’ll know in 14 days. The metric is binary — does the Shipped panel populate for any user? If yes, the UX bet was right and I extend the pattern. If it stays empty, the diagnostic was wrong and the gap is in the thesis, not the UX. Different pivot.

I’ll post the results when the window closes.

Where this came from

I’m a developer at a Fortune 500 by day and a solo founder by night. aicofounders.co is what I’m building — AI cofounders for solo founders who want to ship a SaaS while holding down a full-time job and a family.

The product is in closed beta. If you want to try it, join the waitlist. If you want to test it without signing up, you can run a free idea teardown — one of the cofounders, running for free, on whatever startup idea you’re thinking about.

If you’ve solved the engage-but-don’t-ship gap in your own product, I’d love to hear what worked. Reply on X or shoot an email — I read everything.

I'm building a SaaS at night while working at a Fortune 500 and raising a kid. Here's the math.

Wed, 03 Jun 2026 00:00:00 GMT

The math nobody publishes

Eight hours at the day job. Six hours awake with my son and partner. Two hours on the side project, 10 PM to midnight, most days. Eight hours sleep on a good week.

That’s the schedule. Most weeks I get 12-15 hours on the product. Some weeks the day job demands a fire and I get five. Some weekends the kid is sick and I get zero.

At 50 hours a month, that’s 600 hours a year. Most full-time founders burn 600 hours in a single month. The math says I should be 12x behind them.

Six months in: I’m not. Slower, yes. But shipping consistently, learning the same lessons (just slower), and — crucially — not running out of money or burning out.

I want to share what actually works at this pace because most “founder content” is written by people who quit their jobs. The side-hustle path is undermarketed.

Why I haven’t quit yet

Three reasons.

Stability removes desperation. I’m not building from financial fear. I can pick the right feature instead of the marketable one. I can spend a week fixing infrastructure that won’t show up in screenshots because there’s no urgency to fake progress. When you’re not running out of money, you can make slow decisions.

The day job teaches what NOT to do. I sit in 4-hour quarterly planning meetings. I watch products take 3 months to ship a button color. The contrast trains my speed instincts. Every time I open my laptop at 10 PM, I can ship in 25 minutes what a corporate team would schedule for next sprint.

The audience is built-in. My LinkedIn audience cares about “engineer at Fortune 500 builds side project” content. That’s the content I have. Doesn’t work without the day job context.

The conventional wisdom says quit, give it everything, the constraint of survival will force product-market fit. My experience: the constraint of survival forces something, but it’s often not PMF. It’s bad fundraising, bad customers, or burnout. Stability lets you wait for the right shot.

The actual schedule

Monday-Friday:

7:00 AM — kid up, breakfast
8:30 AM — at the day job
6:00 PM — back home, family time
9:00 PM — kid in bed
10:00 PM — laptop open
12:00 AM — laptop closed (mostly)

Saturday-Sunday:

One of these days is family-only, no laptop
The other has maybe 3-4 hours of focused work
Weekend hours are unreliable — plan around weekdays

That’s it. The 10 PM block is sacred. No phone. No Slack. No “quick check” on email. Two hours of clean coding.

The 2-hour evening block — what actually fits in it

I used to think 2 hours wasn’t enough to do anything real. Six months in, here’s what fits:

One bug fixed + one feature shipped
One new tool wired through the integration layer
One blog post written
One round of cold emails (5-10) personalized and sent
One full code review of my own work from the last month
One conversation with a beta tester (DM thread)
One Notion template written from scratch

What does NOT fit:

Two of the above
A 4-hour deep refactor
A real podcast appearance
A live coding session

The discipline isn’t “work harder.” The discipline is “do exactly one thing per evening.” Pick before you open the laptop. Don’t context-switch.

What I stopped doing

A list of things I stopped doing once I accepted the 12-hour-week constraint:

Reading hacker news during work hours. I used to scroll for 20 minutes a day “for inspiration.” Now I batch it to weekends.
Tracking every metric daily. Daily metric reviews are a procrastination loop. I look at numbers once a week.
Writing roadmap docs. Nobody reads them. The work itself is the spec.
Adding features users haven’t asked for. Easier said than done.
Meetings with myself. I used to “plan the week” for 45 minutes on Sunday night. Now I write the weekly ship target on a Post-it, that’s it.
Premature scale thinking. Optimizing for 10,000 users when I have 2 is just procrastination dressed up.
Comparing myself to full-time founders. Their constraints aren’t mine.

What I started doing

One ship target per week, written before Monday. Anchors everything.
One personal email per day to a beta tester or specific ICP person. 30 emails a month, slowly compounding.
One blog post per week (this is one). SEO compounds over 18 months.
One free resource per month (Notion templates, frameworks). Lead magnets that demonstrate the product.
A weekly “what I shipped” review every Friday night, 15 minutes. Brutal honest.
Walking before the 10 PM block to context-switch out of day-job brain.
A list of “would do if I had a full-time founder budget” that I check quarterly. Most things on it turn out to not matter.

The 6-month numbers

Side-hustle math, for context:

Followers grown: 90 → 370 on X
Waitlist signups: 92 (all organic)
Beta invites sent: 18
Beta accounts redeemed: 3
Active users: 2
Real customers (paid): 0 (still closed beta)
Hours spent: ~350
Hours per beta account: 116
Hours per X follower: 1.25

The hours-per-result math is awful. The compounding math is fine. The fundamentals (shipping consistently, talking to users, learning the lessons) are working. The amplification (distribution) is where the slow part is.

What I’d tell my 6-months-ago self

You’re not behind. The pace is normal. Founders you admire took 1-3 years before anything worked.
Distribution is the bottleneck, not building. Spend 50% of your time on distribution starting now. (I still don’t do this enough.)
Pick a narrower niche. “Solo founders” is too wide. Pick a specific subgroup — side-hustle founders with day jobs, or domain experts going SaaS — and write FOR them, not for everyone.
Build a free vehicle product early. A small free tool that pulls people into the funnel without committing to a SaaS trial. (I just shipped one — should have done it month 1.)
Stop “saving” Product Hunt for “later”. Launch it on a small product to learn the mechanics. The first launch is always rough.
The reply game is your most underused channel. Bigger account threads are where attention lives. Be there daily.

Why I’m still doing this

The honest answer: it’s the most meaningful work I do in any given week. The day job pays well and I’m grateful for it. But the day job hasn’t asked me to think hard in months. The side project asks me to think hard every single night.

That’s worth a lot. Even if the numbers stay slow, the I’m a person who builds things identity stays alive. I can’t put a dollar amount on that.

The day job and the side project teach each other. The corporate context grounds me in scale and stakeholder reality. The side project keeps me dangerous.

Where to find the work

I’m building aicofounders.co — 6 AI cofounders for solo founders who want a team without hiring one. Closed beta.

If you want to test it without signing up, run a free idea teardown. It’s the Product cofounder, running for free, on whatever startup idea you’re thinking about.

If you’re on the same path — day job + side project + kid or no kid — say hi. I read every email and reply within 24 hours: kyle@aicofounders.co.

This is post #2 of an ongoing build-in-public series. Subscribe to follow the journey from closed beta to first paid customer.

My LLM cost was 3x wrong for two months. Audit your own dashboard.

Wed, 03 Jun 2026 00:00:00 GMT

The trigger

I run a SaaS where every user interaction triggers LLM calls. The product is aicofounders.co — 6 AI cofounders that produce strategic output for solo founders. Token spend is the largest variable cost in the business.

Two months ago I built a per-user cost dashboard for the admin panel. Looked clean. Showed me $4.50/week in LLM cost. Felt expensive for 2 active users.

Made me nervous. Started slowing beta invites because the math felt fragile. Considered raising the Starter tier price from $29 to $39 to cover real cost. Deprioritized some token-heavy features.

Last week I looked at the dashboard again and noticed something. OpenRouter — my LLM provider router — returns response.usage.cost on every API call. The actual cost in USD, calculated by them, returned in the response.

I’d been ignoring it. My dashboard was using a static pricing table I’d built months ago, calculating cost from token counts. I wired up the real cost reading.

The corrected number for that same week: $1.50.

Three times less than I thought.

What went wrong

The static pricing table had this structure:

const PRICING = [
  { prefix: "anthropic/claude-haiku-4-5", input: 1, output: 5 },
  { prefix: "anthropic/", input: 3, output: 15 }, // catch-all
];

For two months, my product had been silently using anthropic/claude-haiku-4-5 (cheap, $1/$5 per million tokens). My pricing table HAD that specific entry — but a typo in the model name. The real model returned by OpenRouter is sometimes served as anthropic/claude-4.5-haiku (version before tier, different convention).

My exact-match check failed. Costs fell through to the catch-all anthropic/* which pointed to Sonnet pricing ($3/$15 per million). Three times the real cost.

The bug was invisible because:

The dashboard “worked” — it showed numbers
The numbers were plausible — $4.50/week feels like a reasonable LLM cost
There was no error log because the catch-all matched correctly (just with wrong rates)

This is the most dangerous category of bug. Loud bugs you catch immediately. Silent bugs that produce plausible-but-wrong numbers can run for months.

Strategic decisions I made on bad data

In the two months when my dashboard was lying:

I slowed beta invites. Was worried about cost-per-user. Should have invited more aggressively.
I considered raising Starter pricing $10/month. Would have hurt conversion for no real reason.
I deprioritized features that consumed tokens. Specifically, the deeper research tools (multi-step reasoning, longer context). Those should have stayed.
I planned a tool-model swap to reduce cost on what I thought was the second-biggest token consumer. Turned out it wasn’t.

Four bad decisions, partially mitigated by being a careful person, but the bias was real. Bad dashboard data shapes product strategy in ways that look rational from the inside.

The fix

Two parts.

Part 1: read the real cost.

export function extractUsage(responseData: any) {
  const usage = responseData?.usage;
  if (!usage) return null;
  const rawCost =
    typeof usage.cost === "number"
      ? usage.cost
      : typeof usage.total_cost === "number"
        ? usage.total_cost
        : null;
  const cost =
    typeof rawCost === "number" && Number.isFinite(rawCost) && rawCost >= 0
      ? rawCost
      : null;
  return {
    promptTokens: usage.prompt_tokens || 0,
    completionTokens: usage.completion_tokens || 0,
    totalTokens: usage.total_tokens || 0,
    cost,
    servedModel: typeof responseData?.model === "string" ? responseData.model : null,
  };
}

When the provider returns real cost, use it. Static table is fallback only.

Part 2: backfill historical data.

Rather than just fixing forward, I wrote a script that walked every UsageLog row in the database, looked up the correct rate for the model recorded, and updated estimatedCost to match reality. About 5,000 rows. Total time: under a minute.

The dashboards now show truth all the way back to the product launch.

A checklist for anyone running LLM-powered SaaS

If you have a per-user cost dashboard and you’ve never re-verified the methodology, do it tonight. Bullet checklist:

Read your provider’s response.usage.cost field if available. OpenRouter, Anthropic, OpenAI all return it. Use it as the source of truth.
Match on both requested AND served model names. The model you REQUEST and the model your provider SERVES can differ in naming conventions. Catch both shapes in your pricing table.
Add an explicit pricing entry for every model variant you use. Don’t trust catch-alls. When you add a model, add the entry. When you remove a model, remove the entry.
Re-audit your cost methodology quarterly. Models update. Env vars change. Provider pricing changes. Your code didn’t notice.
Track requested model vs served model in your log. Saved me a debug round when I needed to figure out which model was actually doing the work in production.
Backfill historical rows when you fix the methodology. Otherwise your dashboards show two different truths — old wrong, new right — and your eyes will lie to you about trends.
Cross-check by hand monthly. Pick a single user, look at their token spend in the dashboard, manually multiply by current rates, compare. Catches drift before it becomes habit.

What this cost me

Two months of slightly suboptimal decisions. No actual lost money — the real cost was always lower than I thought. But the opportunity cost of slowing beta invites and dragging on pricing decisions: probably one or two beta users who would have been ready to be paid converters by now.

The lesson generalizes beyond LLM cost. Dashboards lie. Especially the ones you built yourself. Especially the ones you stopped checking the math on.

Bigger principle

A founder I respect once told me: half your job is being skeptical of your own data.

If you can’t reproduce the dashboard math by hand on a single row, you’re flying blind.

If you haven’t re-verified the methodology in 90 days, the methodology is probably wrong.

If your strategy depends on a number in a dashboard, audit the number before you commit the strategy.

Where this work happened

I’m building aicofounders.co — 6 AI cofounders for solo founders. The cost dashboard described here is part of the admin panel that lets me run unit economics in real time.

Closed beta. Free idea teardown (no signup): aicofounders.co/teardown.

If you’ve caught your own silent dashboard bug, I’d love to hear the story. Reply on X or email kyle@aicofounders.co.