I’m excited to announce that @humanicai has been acquired!
We built Humanic to capture how the best sales teams think, decide, and act to translate that into actionable context. Enterprise sales conversations deserved a true decision layer for automated judgment, and after 2M+ minutes of deal conversations, we’re proud to have set that in motion.
Growing up in India, @abhi_is_agi and I admired the behemoths of Silicon Valley, but joining forces with one was not on our cards. It’s an honor. I wish I could divulge more, but our lawyers say we have to wait.
We want to thank our earliest believers, our users, and the ones who stayed. It's still day one here in SF, and we are just getting started!
See you in the next chapter.
Fable’s ban shows that models will start getting restricted.
Agent harnesses will bear more of the weight. They are becoming their own optimization surface.
Good read from HarnessX.
I built a tiny coding-agent harness and put it head-to-head with OpenCode. Same model, 3 benchmarks, 2 model tiers.
Instrumentation beat intuition.
Kavel used fewer tokens per solve in all 6 matchups.
Polyglot/36: both 100% pass, Kavel at 54% of the tokens.
Swapped in a reasoning model with zero retuning → still 100%, on even fewer tokens. 🧵1/n
My wife is running an event in San Francisco today for people who throw events.
So I have a free day in San Francisco that I didn't schedule up.
Any entrepreneurs want to meet up to show me what you are working on?
We interviewed with YC for S26.
Those 10 minutes completely reshaped what we’re building. I’ve gotten good advice over the years, but YC feedback is next level; diagnostic, not prescriptive. Strongly recommend.
Since then: signed 3x design partners, doubled revenue, and expanded to 30% more use cases. We will be back for YC.
If capturing valuable tacit knowledge in your org sounds exciting, hit us with your use case 👀
AI won’t replace humans.
It’ll just replace juniors, rename seniors “agent managers,” and let CEOs call productivity has entered a new paradigm.
In this process your org never learns
@contextconor@garrytan@t_blom@JayaGup10 Don’t you think it’s hope chess? Companies can’t see the effect immediately, why’d they invest on something that builds up instead of showing up on day 1?
OH in SF: tacit data in cos are a 80/20 problem - AI can automate most usecases , you need implicit data for the rest 20%. Won't pay for that little effect.
Maybe true, that breaks down a lot of sf assumptions about this space.
i want a thing that can just ask me 'yo whatsup, why'd you do that thing?' and go about it's day. It keeps learning, my life gets easier.
magic happens in the backend. minimal data extraction as a moat.
nowadays, the assumption is that you can just infer and collect "tacit" knowledge by scraping artifacts like slack messages, emails, docs, etc. But that's often insufficient.
If it's written down in those artifacts, it's no longer tacit or implicit... it's explicit.
I don’t go around documenting the real why behind my decisions in slack or email. That reasoning lives in my head. So if you want actual tacit knowledge, you have to capture it from the source (the person) directly, not try to reverse-engineer it from the artifacts they left behind.
for example, if i need to know something, I don’t keep talking to the company RAG, that already has access to all the artifacts, but I go and ask the person why.