
hunether
1.6K posts



Hot from the lab: Google shipped Gemini 3.5 Flash a few hours ago. SERV engine instantly made it better. Deeper cut: SERV + DeepSeek v4 Flash still beats Google at 1/30th the cost. SERV makes general models production-ready for Fortune 500s, governments, and high-stakes ops.





Gonna flip my $REI into $SERV — here’s why 👇 @rei_labs is trying to make LLMs work differently at the core layer. @openservai is doing something more immediately useful: It makes existing LLMs more efficient, reliable and transparent through Serv Reasoning / BRAID.




Congrats to everyone who got in to $TSG but I have to say it's incredibly blackpilling for onchain to be revived by strong AI utility plays... Only for base and the majority of KOLs to start pushing a shitty meme again. Luckily I derisked before all this but it appears the onchain community will never learn from its mistakes.

Thank you to someone unbiased and with common sense Ffs, never seen this amount of cope and losers ever on my timeline "How dare base support a meme, this chain is for ONLY fundamentals, you have to ONLY support the fundamentals' $TSG literally the most organic meme i've seen and we have people trying to bring it down for no reasons. Cringe af







$SERV | @openservai is building a v serious product but no one seems to pay attention (yet) $SERV is turning SERV Reasoning into one of their core products, making LLMs smarter and cheaper Some numbers that surprised me recently: - Grok 4.3 + SERV Reasoning beats vanilla Grok 4.3 and GPT-5.5 - DeepSeek v4 flash + SERV = 36x performance-per-dollar vs GPT-5.5 Not many people seem to be watching, but if you spend 10 minutes on their socials the strides are obvious - Apr 11-15: Beta waitlist push, 200+ applications - Apr 15: Reasoning API enters private beta w governments running it in production + enterprise pipeline growing - May 1: First beta cohort skewed to real world use cases (genomics, pentesting, satellite imagery, edge AI..) - May 8: 29,638 requests + 279M tokens in beta wk1, waitlist packed, batch 2 invites going out soon Given how consistently they're shipping and that Reasoning is about to get more accessible, I think they can do well from here Btw, whatever model you use works with Reasoning, all OpenRouter models Last but not least, I like how this chart is starting to look.

spent a lot of time with the @openservai team over the last 18 months very few teams have the vision and the execution to find PMF in an unfathomably large emerging TAM - AND be positioned for it at the right time truly one of a kind. excited to be going on the journey. full 22/7 thesis here: 227.fund/blog/openserv-…






