Founder @ Real Good GTM. i build GTM systems and scale pipeline for B2B companies. scaled palm ai, cef ai and Boolee before this as first business hire
everyone on my timeline has been talking about Instinct. an AI assistant inside WhatsApp that just does things.
decided to test it with something small.
"book cafe peck koramangala, 2 ppl, tomorrow 7:30, cheapest platform"
it found the app with 10% cashback, told me the cancellation terms, asked for my OTP, and confirmed the table. all inside the chat.
ok. i get the hype now.
@noahrshinn
ps. I have few invites left if want to try it out
Gave TypeSafe's Jev the test i'd give a new SDR.
100 made-up companies, 63 fields each, my real ICP rules. Fit or not, which tier, why not.
It got all 37 disqualifications right, with the reason. Pushed zero junk through. Whole run cost me about a cent. Same tokens through Claude or GPT would be dollars, plus me fishing JSON out of a paragraph (again).
That's the part i care about, not the cent. Claude and GPT hand you text and a confidence they made up on the spot. Jev hands you a probability that actually holds, a pick from your list, or a score on your scale. Nothing else can come back. There's no paragraph for it to hide in.
And no, i didn't move to a new platform. My Claude agent still runs the whole thing, it just calls Jev for the yes/no instead of thinking out loud about it.
Where i'd use it: killing junk off a raw list before Clay spends a credit, sorting replies, deciding if a job post is a real signal. Anything fuzzy, anything that needs words, Claude still does.
@typesafeai
Gave TypeSafe's Jev the test i'd give a new SDR.
100 made-up companies, 63 fields each, my real ICP rules. Fit or not, which tier, why not.
It got all 37 disqualifications right, with the reason. Pushed zero junk through. Whole run cost me about a cent. Same tokens through Claude or GPT would be dollars, plus me fishing JSON out of a paragraph (again).
That's the part i care about, not the cent. Claude and GPT hand you text and a confidence they made up on the spot. Jev hands you a probability that actually holds, a pick from your list, or a score on your scale. Nothing else can come back. There's no paragraph for it to hide in.
And no, i didn't move to a new platform. My Claude agent still runs the whole thing, it just calls Jev for the yes/no instead of thinking out loud about it.
Where i'd use it: killing junk off a raw list before Clay spends a credit, sorting replies, deciding if a job post is a real signal. Anything fuzzy, anything that needs words, Claude still does.
@typesafeai
What's actually working in GTM right now:
→ Founder-led content on LinkedIn or X. Genuine POV, 3x a week. Still the highest ROI channel at early stage.
→ Small, weird communities. A 200-person Slack group mention converts better than a 50K impression ad. Find where buyers actually hang out.
→ Outbound that doesn't look like outbound. Cold email is dying. Warm context wins - engage someone's content for two weeks before you DM them.
→ Ungated everything. The gated PDF is dead. Give away the frameworks, sell the implementation.
→ Dark social proof. Happy Slack screenshots, lo-fi customer quotes, a random tweet. Polished case studies aren't hitting like they used to.
What did I miss?
A model built for machines, not chat, with a zero hallucination claim. @typesafeai
Skeptical by default, but this is the first launch in a while aimed at the actual bottleneck for agents in production: not reasoning, not speed, just not making things up.
After co-inventing ChatGPT, I kept asking myself: why have superhuman chat models not led to AGI?
I’ve spent the last 2 years in stealth building a new way to train models (RLCD), and a new type of frontier AI model that we are releasing today: Jev
• 20-200x faster
• 40-400x cheaper (w/ output tokens free)
• Frontier composable intelligence optimized for decisions
AFAICT the shortest path to AI-based economic revolution