After working in crypto DevRel for half a decade, I have a confession I hate to admit.
I think DevRels are one of the main reasons why crypto hasn't reached mainstream adoption.
Here's what's going on:
I joined Y Combinator with 14 users, building a dev tool for myself.
Then @SaaiArora joined as CTO, and we started growing and shipping 2x faster. It forced us to rebuild product and infra from the ground up for teams, not solo devs.
We are ending the batch with >300k agent minutes per week, and teams like @mintlify, @knowunity, and @modaflows shipping through Replicas every day.
Today we're launching Replicas V2.
I'd love to be able to save pinned answers directly from a @claudeai conversation to go back to them quicker in the future.
I'm surprised that's not yet a feature.
"Teach me to invest and I'll teach you how to code" - I told my crush back in 2019.
7 years later - not only did I get a husband out of that date, but I nerded over markets so hard, my girlfriends started urging me to build them an investing course.
I kept saying "I don't have the time".
And then we sold Helicone in February. I spun up a quick Luma site, dropped it in a few Whatsapp groups, and launched the first cohort of "Wealthy Women Don't Wait".
20+ women joined in less than 15 days!
Then, a family office reached out: they wanted to offer the course to their members. Cohort 2 got a date, a curriculum & tech stack upgrade, and a higher price tag.
This time, 25+ women signed up (for almost double the price!).
We started Cohort 2 this Monday and the one quote stuck in my head is:
"Money loves speed. Poverty loves waiting."
~ Alex Hormozi
Life rewards those with the audacity to jump on opportunities. I have a feeling this is just the beginning 🎢.
Today, we’re launching the @link wallet for agents. It lets you securely empower agents to spend on your behalf. Your payment credentials are never exposed and you approve every purchase.
https://t.co/TcvEiVNth9
Over the last week, we've automated everything except humans clicking the merge PR button.
Introducing Replicas Automations:
- daily security reviews / automated pentests
- auto-fix customer Slack requests silently
- auto-fix sentry / datadog errors
- PR description and Linear sync
- much more
Since we began using it internally, we've been able to automate out an absurd amount of the SDLC.
1. On-call engineering
Replicas automatically responds to errors in our Datadog. Additionally, every Slack customer message is silently picked up by Replicas, where it will evaluate, write a PR, and ping me in a different channel.
2. Code Review
Since Replicas resolves CI failures automatically, we tracked the 8 most common things we typically rejected coding agent PRs on (useEffects, WET code, doc updates).
We made 6 automations to cover them, all of which instructing Replicas to make CI checks.
By combining this with reviewers like @greptile, Replicas writes the PR, then iterates over and over until all the OTHER Replicas agents agree that it passes.
By the time any engineer reads the PR, its nearly perfect.
3. Testing
Every day, Replicas runs E2E tests on our public API. For any unexpected failures, we get a Slack message.
4. Planning and Development (duh)
Replicas already had integrations with Slack and Linear, so developing with it was naturally collaborative.
Many of our beta testing organizations have found incredible use cases, like full E2E fullstack tests, pentesting, and much more. And we have many exciting triggers and automations to come.
Book a demo with us for replicas[dot]dev and I promise we can make engineering 2x faster ;)
Sequoia's thesis that the next $1T company will sell work, not software, is the most important reframe in AI right now.
The argument: if you sell a copilot, you're competing with every new model release. But if you sell the outcome — books closed, contracts reviewed, claims handled — every AI improvement makes your margins better, not your product obsolete.
The key insight most people miss: for every $1 spent on software, ~$6 is spent on services.
The entire SaaS playbook was about capturing the software dollar. The AI playbook is about capturing the services dollar — at software margins.
Not "AI for accountants." The AI accounting firm.
Not "AI for lawyers." The AI law firm.
The companies that figure this out won't look like SaaS companies. They'll look like services firms rebuilt on software infrastructure.
That's a fundamentally different company to build, fund, and scale. And most founders are still building copilots.
call me crazy but I think SaaS will slowly disappear
Im seeing more and more non-technical folks ship SaaS-like products for their teams in a few hours.
why use someone else's product when you can build one for your own needs and maintain it with a long-running agent?
we're at the point of the cycle where "just bc we can, doesn't mean we should"
just bc AI can do it, doesn't mean it should
just bc we can tokenize it, doesn't mean we should
just bc it can be a mobile app, doesn't mean it should
I call it the boomerang moment
guys call me crazy, but Ive been enjoying codex more than claude code lately?
claude code generates better code on the first go, but codex is just better at debugging - which, lezzzzbehonest, is half the job