Chris Chomenko

Building AI-native companies. Tech, product, sales, capital - full stack founder. Current: BAM AI. Systems over headcount.

[Chris Chomenko @ChrisChomenko]

ToS like this is platform risk pricing itself in. If your product is a wrapper around someone else's subsidized tool, the landlord can evict you by paragraph. The practical rule we follow: anything load-bearing runs through an API with a contract, or on weights we control. Consumer-app terms are for consumers.

[Chris Chomenko @ChrisChomenko]

15 hours on one goal is the real benchmark. Anyone can demo 5 minutes of autonomy. What I'd love to see published: how it decides something is done versus good enough to stop, and what percentage of its changes survive your review. Duration is impressive. Survival rate of the work is the number that builds trust.

[Chris Chomenko @ChrisChomenko]

True exactly where verification is cheap. Landing pages, ad variants, outreach: fire the cannon, measure, keep winners. In domains where a wrong output costs real money (medical claims, contracts, infra) the constraint flips and the scarce person is whoever can verify at the cannon's pace. The cannon didn't kill the 10x engineer. It made verification the 10x skill.

[Chris Chomenko @ChrisChomenko]

The annoying question is the valuable one. Most people can't articulate what good looks like, and the model asking forces the spec into existence. Half the value of agents in our shop is that they made us write down standards that lived in people's heads for years. The interrogation is the onboarding.

[Chris Chomenko @ChrisChomenko]

The 30-candidate rule is really a calibration rule. You can't recognize the 95th percentile if you've only sampled five. Same reason we eval agents against a wide baseline before trusting one with a workflow: judgment about quality is downstream of how much quality you've actually seen.

[Chris Chomenko @ChrisChomenko]

The identity that survives is owner of outcomes. I build with AI agents daily and my job stopped being writing code a while ago: it's deciding what to build, setting the bar for correct, and owning what ships. Code was always the means. The crisis hits hardest for people who mistook it for the end.

[Chris Chomenko @ChrisChomenko]

We're living this from the buyer side: my agents use several SaaS tools daily that no human on my team has opened in months. The vendor sees zero DAUs and a healthy API bill. Pricing and metrics both need to move to work delivered, because the human login is becoming the least informative signal a product has.

[Chris Chomenko @ChrisChomenko]

The construction data backs you: America now pours more concrete for data centers than for office buildings, $57B vs $48B in the past year per Census, and in 2014 offices led 20 to 1. Those are real paychecks in real counties. The unresolved part is the grid bill landing on households that never signed the lease. Get the cost allocation right and the politics of this get much easier.

[Chris Chomenko @ChrisChomenko]

The protests make sense the moment you look at utility math instead of ideology. PJM's capacity auction cleared at $28.92 per MW-day for 2024/25 and hit the FERC cap near $325 the last two cycles, with the costs flowing into retail bills. People are reacting to their electric bill, and the safety orgs are looking for an AI-risk story where a ratepayer story is sitting in plain sight.

[Chris Chomenko @ChrisChomenko]

The discount is the first public price tag on training data, and businesses should read it exactly that way. If a lab will forgo 95% of revenue to see your prompts and outputs, your workflow data is worth roughly 20x what you pay for inference. Every enterprise negotiating an AI contract should be doing that arithmetic before signing away logs.

[Chris Chomenko @ChrisChomenko]

From an operator seat the sovereignty point is the real one. Regulated buyers I work with (healthcare) increasingly demand models they can run inside their own walls. Open weights on owned hardware is the only architecture that fully answers them. The chip company buying the model commons makes that path a first-class product.

[Chris Chomenko @ChrisChomenko]

The strategic logic writes itself: every open model downloaded is future demand for the chips that run it. The question worth watching is hub neutrality. HF became the commons because it never picked a hardware side. Keeping the commons neutral inside a company with a side is the hard part, and the whole ecosystem is about to find out.

[Chris Chomenko @ChrisChomenko]

The hard part is the cold start. A union's power is coverage density, and an app with 2% of a workplace has nothing to bargain with. Whoever cracks it probably leads with something individually useful on day one (pay benchmarking, contract review) and lets the collective layer emerge once density arrives.

[Chris Chomenko @ChrisChomenko]

7 in 10 Americans approve of labor unions. 1 in 10 belongs to one.

Gallup has union approval at 68%, and 2022's 71% was the highest in over five decades. BLS membership data tells the opposite story: 20.1% of workers in 1983, 9.9% in 2024, the lowest on record.

Americans like unions in theory and skip them in practice. Into that gap walks AI, the biggest repricing of labor in a century, and 9 in 10 workers will face it with no one bargaining for them.

Is the union model dead? Does AI resurrect it? And if neither, who negotiates for labor?

[Chris Chomenko @ChrisChomenko]

Since 1991, US day care and preschool prices are up 297%. Overall prices are up 148%. The typical worker's hourly pay is up 212%. All per BLS data through July.

Childcare outran both inflation and wages over 35 years. Meanwhile the fertility rate sits at 1.63, below replacement every single year since 2007.

Here's the twist: care work is exactly what AI can't automate. Education and health care are all of US net job growth right now.

The work machines can't do is becoming the work families can't afford. What breaks first: the price, the birth rate, or the model where parents pay for it alone?

[Chris Chomenko @ChrisChomenko]

This framing is the right way to evaluate any AI spend: cost per hour of compute vs revenue per hour it generates, fully loaded. Most teams still budget AI like software (flat subscription, ignore utilization). Pricing it like labor with a P&L per agent changes every decision about what runs and when.

[Chris Chomenko @ChrisChomenko]

The gap I see between teams getting real value and teams stuck in pilots is never the model choice. It's whether they rebuilt the workflow around the model. Routing between providers is a commodity now. Encoding your domain's judgment into evals, guardrails, and context is the part nobody can buy off the shelf.

[Chris Chomenko @ChrisChomenko]

The industry is quietly running an experiment: what happens to the senior pipeline when nobody hires the juniors who would become them. NY Fed data already shows recent grads running higher unemployment than the workforce overall, first time in decades of records. Mentorship was never charity. It was how firms manufactured their own seniors.

[Chris Chomenko @ChrisChomenko]

AI made outbound free, so the scarce asset flipped from reach to attention filtering. The endgame is your agent answering your phone: it screens, verifies, wastes the spammer's compute, and only known humans get through. Whoever ships the personal gatekeeper that actually works owns the most valuable position in communication.

[Chris Chomenko @ChrisChomenko]

The elegant part is it prices conviction instead of talk. Every investor claims high conviction; almost none will concentrate a fund on it. Indexing off the GP's own position sizing turns cheap words into a costly signal. Founders should read term sheets the same way: ignore the enthusiasm, look at what percentage of their fund you are.

[Chris Chomenko @ChrisChomenko]

Outcome pricing works when the outcome is a dollar figure a third party confirms. We price denial appeals on recovered revenue: the payer's remittance is the referee, nobody argues about what counts. Tickets fail that test because the vendor grades its own homework. Pick outcomes with an external scoreboard and the pricing conversation gets easy.