Some of the biggest companies of the next decade won't be software businesses. They'll be services companies like insurance carriers, law firms, and tax practices rebuilt from scratch with AI doing most of the work.
In this episode of Startup School, YC Visiting Partner @CharlieWarren walks through the playbook for building AI native services companies, covering how to pick a market with the right traits, why variance kills these businesses faster than anything else, and the P&L math that’ll transform your business model.
00:00 — Intro to AI Services Companies
01:01 — Picking the Right Market
02:55 — Markets YC Likes Right Now
03:43 — The Sam Altman Test
04:35 — The Right Founding Team
05:28 — Building the Product
06:19 — Variance Is the Existential Problem
07:08 — The Early Demand Trap
07:53 — How to Price AI Services
08:41 — The P&L Walkthrough
09:33 — AI Operating Leverage
10:27 — Don't Buy Your Way In
You've been asking for this one...
Now in preview: Codex in the ChatGPT mobile app.
Start new work, review outputs, steer execution, and approve next steps, all from the ChatGPT mobile app. Codex will keep running on your laptop, Mac mini, or devbox.
@michael_lwy@NousResearch for your cases, github is enough, just rent VPS on @zeaburapp, and create a github account for your agent, grant the write access to your repo, every time it did something just git pull the latest repo
Today marks a major milestone for KAIO with the $KAIO TGE.
A quick reflection on where we stand after 2 years of building:
• ~$100M TVL live with tokenised top-tier funds from BlackRock, Brevan Howard, Hamilton Lane, Laser Digital, and Mubadala Capital.
• Live across 10+ Tier-1 chains with our cross-chain gateway including @solana, @SeiNetwork, @SuiNetwork, @Aptos , @hedera, and others.
• Backed by top tier institutions who are rewriting how capital markets and money moves including @tether, @LaserDigital_, @BHDigitalAssets, @Systemic_VC, @LyrikVentures, @Karatage_, @ShorooqPartners, and others.
• Building toward deeper DeFi composability that bring real utility to RWAs
• KASH, our retail RWA access product, is launching soon to open up even broader access.
The tokenised asset space is hitting a real inflection point. For KAIO it has always been traction and product before hype. We’re here to lead this transition to onchain capital markets. This is what we mean by Transforming Institutional Funds Onchain.
This is close to what I was hoping to communicate with https://t.co/mb2DyT1eX4
A lot of people still carry a fixed idea of what issue tracking is: a place to record work, assign it, queue it, and report on it after the fact. That mental model shapes how they use the tool.
The landscape has changed, but the mental model often has not. From the old perspective, it can look like issues no longer fit the way work happens. Work moves into agent sessions, Slack, code, and PRs.
We saw a version of this before AI. Teams would move from Jira to Linear, but still expect Linear to behave like Jira, even though we intentionally designed it around a different way of working.
Now the shift is that the issue is no longer just a record of work, or a queue for future execution. It can become the interface for directing, observing, and coordinating work as it happens.
I do most of my AI-assisted work from the context that already exists in Linear: customer feedback, product discussions, planning, projects, and prior decisions.
That context helps the agent understand the work better. From there, I can launch issues or projects for the agent to work on, then review, steer, and verify the result in the same shared environment.
Context → execution → verification → loop back.
That is the new mental model: Linear is where you build shared context, launch execution from that context (even if sometimes the work is launched to local agents or other systems), track the work, verify the outcome, and preserve the loop.
Our challenge is to communicate that shift more clearly, and to make the product itself support and encourage the new behavior.
Goblin and related magical mentions were overrewarded in training, and the behavior was reinforced over successive models.
We removed the goblin-affine reward signal for future models, and filtered training data where creatures appeared in irrelevant contexts.