LFM 2.5 (with some mods) has been our go-to since we entered regulated/air-gapped verticals. The level of brilliance from Ramin, Matthias, and co cannot be overstated.
some vc bros had a recent epiphany and finally discovered the urgent necessity and unlimited tam of on-device/local ai after like 4 years, speaking about it like this is the radical new frontier they introduced to the world just now. like bruh where’ve you been consensusmaxxing!
lag. it is fascinating to me how a majority of tech vcs operate by solo pattern matching with a recency bias towards what they hear from their gods.
For scoped tasks in generalized domains—yes, but otherwise fable 5 on ultracode with workflows was 👨🍳💋, night and day difference for greater depth+complexity codebases. Shines in distributed systems and can drive itself (completely autonomously when paired with memory), unlike opus and gpt.
@NetworkChuck Huge gap, considering how many industries are required to be air gapped, and how far behind govcloud is from frontier models for controlled data. We’ve been working on squeezing out performance from SLMs and other local models. Excited to see what everyone else is building!
The US government, citing national security authorities, has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees.
The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance.
Access to all other Claude models is not affected.
We apologize for this disruption to our customers. We believe this is a misunderstanding and are working to restore access as soon as possible.
Read our full statement: https://t.co/bwn0sximKZ
@Brennan_Lup building https://t.co/CvnBpydvXY half of our stuff is OSS, we just finished compliance and offline/static platform w/local models for deftech/aero sub primes. Diving into RL for SLMs.
@CiaraACade Saw your earlier work, compliance is a great space, why the pivot? CMMC level 2 deadline is looming in deftech and there's hella alpha abound using your existing tech
Claude Fable 5 is amazing at most of our team's technical tasks, but the moment it sees a *bad word* like AES/AVX or biology/drug it nukes itself back to opus. It's impossible to use for frontier/research or ML tasks.
the cost of shipping code went to zero
taste didn't
but "taste" sounds mystical and unfixable, so nobody teaches it. here's the unmystical version: taste is just an eval you haven't written down yet
how you choose what to measure is what matters
1/8
"You need to make yourself a big target for luck, and the way to do that is to be curious. Try lots of things, meet lots of people, read lots of books, ask lots of questions."
— Paul Graham, How to Do Great Work
https://t.co/AyIk5fnKZL
@zacharyvalles if you guys just need a crash pad, https://t.co/9GPURvBZlp might be the spot, otherwise bookface is always a good resource for P26 folks heading out!
@dewanshranjan@ycombinator IMO, it's worth it. The feedback that you get is gold. As a sidenote, your product directly addresses some of the points I get from senior engineers about their fears w/r/t AI tooling compressing cycles at the workplace. Cool stuff guys!
We're hosting an event on June 16th in San Francisco.
Compile is a one-day event that brings together engineers, researchers, designers, and builders of all kinds to discuss the future of software.
https://t.co/8YERlPFooL
sv people underestimate operational complexity in healthcare
healthcare people underestimate ai leverage
so what do you get?
copilots that no one wants