vineetmehta

building self-learning agents + growth loops, minus the hand-waving

yc founder, ex-opendoor, ex-vc

[vineetmehta @vineetmehta]

@arian_ghashghai what's your split of investments that came inbound (like what your post outlines) vs outbound where your team reached out?

curious because the @ycombinator playbook of posting progress publicly brought me enough inbound that we closed our round before demo day (few yrs ago)

[vineetmehta @vineetmehta]

@heyalexmoore this is fun. could be an agent that runs everyday!

[vineetmehta @vineetmehta]

@AGrillz every batch, the bar rises 🚀

[vineetmehta @vineetmehta]

@mikenevermiss hard to un-see this as the future of software🔥

[rahul @rahulgs]

1. as a mental model it is more correct to think of fable+ class models as english -> code interpreters - converts your idea into code into "correct" code regardless of problem complexity and output complexity (diff size). Fable 5 will be the worst of this new class of models

2. diff size/complexity is to be managed purely for review:
small diffs - in high risk areas of code (auth/identity/data access/network access/money movement)
large diffs for code that can be empirically verified (frontend/backend plumbing/code without network or db access/performance code that can be empirically verified)

3. time it takes to ship software is completely disconnected from time to produce the PR - how long the work takes depends fully on ability to review/merge code while managing risk at scale

4. solving the bottlenecks for above matter enormously- linters/testing/CI/shadow mode verification/empirical verification

5. agency matters enormously- what are the biggest bottlenecks to speeding up the loop and eliminating them? what are the problems that need solving and when do they need solving? what does it take to the solution to all of them today?

6. deep understanding of the full stack matters enormously- what problems are worth pursuing? is there a higher level of problem abstraction to address first? should I give it the sub-sub task, the sub task, or the task itself. what are the major risks with this PR (order of importance: security holes/correctness holes/performance holes). is there a higher speed way of producing data that allows me to merge this? should this be run in shadow or in a sandbox or a flag. understanding every line of logic may not be needed but understanding and managing risk matters enormously.

7. the cost of complexity itself is changing. it might be now worth "maintaining" 50% more code to get a 5% performance win. getting the right abstractions matter less because larger refactors are less tedious. code quality nits become huge drag. very likely, a much smarter model will be maintaining your code so worth taking on more technical debt now. taking the time to hand architect and rebuild systems comes with an enormous cost of velocity

8. if it quacks like a duck and walks like a duck, it's a duck. For low risk cases, it might be more sane to treat code chunks (services / functions) as a black box, like we do for neural networks: do full empirical verification only: has code produced correct outputs for the last 10,100,1000,10k inputs ? can we quarantine this large piece of code - no outbound access to network / database ? what happens when this code is wrong? do we get hacked/or crash(memory/cpu)/is an inconvenience? is it internal facing or external? what can we do to address these risks?

9. eventually, logical verification (line by line review) will come at an enormous cost- save it for where it matters and build systems that are tolerant to empirical verification. is there a decorator that prevents db / network access? correctness bugs are significantly easier to rectify than access bugs

10. what are the rails that allow for even faster iteration? code permissions can be opt in - db writes, db reads, network egress (to where?), PII access. how long does it take to get shadow mode data? how many PRs can be tested? What are the categories of diffs

[vineetmehta @vineetmehta]

The result is that what used to be exceptional startup growth is becoming increasingly common.

The bar is moving faster than most people realize.

[vineetmehta @vineetmehta]

Growth rates are exploding in the latest Y Combinator batches.

[Tim Suzman @TimSuzman]

Tasklet (@TaskletAI) went from $300k revenue at the beginning of the year to $7M revenue run rate at YC Demo Day.

Highest revenue of the P26 batch.

And no coincidence: @startupandrew is a top YC founder, previously founder of Firebase in my S11 batch.

[vineetmehta @vineetmehta]

AI is clearly changing how quickly startups can find product-market fit and scale up.

[vineetmehta @vineetmehta]

7/ Consumer AI gets the headlines. But much of the durable value creation may come from companies going after the unglamorous middle of the org chart, which is where many of the best software businesses have historically been built.

[vineetmehta @vineetmehta]

1/ With YC Demo Day tomorrow, one thing is clear:

Software is becoming labor.

As a YC founder and through my work with @pioneer_fund, I spend a lot of time getting to know each batch. Three patterns stood out to me about Spring '26:

[vineetmehta @vineetmehta]

6/ Put those together and the dominant pattern in this batch becomes clear: a business customer, an agent focused on a specific role, and a product sold as the worker that handles that role's most repetitive tasks.

[vineetmehta @vineetmehta]

We know how to improve cancer outcomes: catch it earlier!

Sadly, MRI diagnostics are limited - too expensive. That's why I'm excited for @pioneer_fund to back founders @ET_adialante & @ManW_dePlan

Their portable MRI that can be brought directly to clinics. Life changing!

[Tim Suzman @TimSuzman]

The Adialante founders are formidable. @ET_adialante and @ManW_dePlan. Will not be stopped.

@pioneer_fund is proud to back them.

They invented and built a new portable MRI and are building a network/marketplace around it.

https://t.co/oEJR5S31hT

[vineetmehta @vineetmehta]

@emollick ironically doctors tend to be averse to using chatgpt in clinical work (see: https://t.co/sK3NQJwDez)

[vineetmehta @vineetmehta]

We’re early in the agent era but one thing is clear: agent-first companies will become as important as mobile-first companies were 15 years ago.
One of many reasons that I'm excited for @pioneer_fund to back @AgentPhoneHQ, which gives every agent a phone number for calls & texts.

[James Fong @jamesfong]

📱 @AgentPhoneHQ gives every AI agent its own phone number, so an agent can actually call, text, and authenticate in the real world. They handle all the telecom compliance behind the scenes, so the agent just signs up over an API and starts making calls and sending messages.

The whole telecom stack was built for humans, which is why agents hit a wall the moment they need to do something as basic as receive a verification code or make a call. Manav and Meet spent months becoming official CPAAS vendors and getting compliant before they ever launched, which is the kind of unglamorous groundwork that turns into a real moat. It's working. Agents are already signing up on their own through the API, and AgentPhone is the default phone number option on Google's Agent Dev Kit, Vercel, Replit, and Langchain. As more agents come online, that distribution compounds.

https://t.co/323VyupN7b

What makes me even more excited is who's building it. @manav2modi and @themeetmodi are brothers who have been building together their whole lives. @pioneer_fund is excited to back them as they give agents a way to reach the rest of us. Congrats!

#yc #ycp26 #vc