kaspar

I do things @featherlessai and @65labslah

[kaspar @aimuggle]

Play with representation engineering from a free tier Colab notebook!

Inspired by @NeelNanda5's TransformerLens notebooks, @ivanleomk and I made a notebook to get up and running with @voooooogel's repeng library in just a few minutes with GPUs available on Colab's free tier.

https://t.co/3gJ9fBCCHF

[kaspar @aimuggle]

@rachpradhan lovingly crafted human slop

[kaspar @aimuggle]

@adlinzainal >(also i think my friends just accept that i am a weirdo)
this tbh

[kaspar @aimuggle]

@agrimsingh just here spreading the good word

[Nick Khami @skeptrune]

"of the course nodes will ship on time, why would i deceive you?"

[Vaughn Tan @vaughn_tan]

Two cohorts done at the non-VC incubator I run in Singapore, a third under way. Same gap every time: startups built to pay for themselves within 12–18 months and then stay small have no funding structure for the stretch before revenue.

We need one.
https://t.co/j9zhkcTfz9

[kaspar @aimuggle]

@AdamHoltererer thats unfortunate - was considering an ultra sub to get access to carbon but might hold off

[kaspar @aimuggle]

@rachpradhan

[kaspar @aimuggle]

@swyxio or opus 5.5 just changed your expectations

[Featherless AI @FeatherlessAI]

Space Bunny is here!

Congrats to @MiniMax_AI on the launch of MiniMax M3.1 Flash 🚀

It’s the best open-weights model on @ArtificialAnlys’ AutomationBench, and it costs a fraction of what closed models charge per task.

Open models keep getting better and cheaper.

The weights drop on Oct 23, live on Featherless the same day:
https://t.co/jASqQqb6P9

[Richelle🚢 @Richelle_Ji]

Introducing Sugar Data.

We asked Step 5 Preview who OpenAI should acquire. It said @duolingo , and made a compelling case (consumer habits, language integration, and ad revenue).

It costs ~1/6 as much per task as Claude Opus 5.5, so we ran it on 100+ more AI companies.

Who's your dada? https://t.co/dt8qmEe1e7

[Featherless AI @FeatherlessAI]

Step 5 Preview is here.

@StepFun_ai's new open-weights model sets a new Pareto frontier, matching Kimi K3 on the @ArtificialAnlys Intelligence Index!

It’s built for long-horizon agent work, with frontier-level results on software engineering and professional knowledge work, especially finance.

- 600B sparse MoE, 27B active per token
- ~1/6 the cost per task of Claude Opus 5.5, per Artificial Analysis

Now on Featherless, served at FP8.
https://t.co/mznI5gFzHe

[kaspar @aimuggle]

@agrimsingh @bot at what cost 😭

[kaspar @aimuggle]

@TheRohanVarma @embirico @gabrielchua also it seems like there are persistent and non-persistent volumes for storage? the main dot vm nuked my custom whatsapp connector that i deployed on it twice. would love to have it all be persistent

[kaspar @aimuggle]

dots are great but the app really needs a dedicated section/project to tuck spawned threads under @embirico @TheRohanVarma

cc @gabrielchua

[kaspar @aimuggle]

@sqs Smells like opus

[kaspar @aimuggle]

accenture has stopped pretending

[kaspar @aimuggle]

@ivanleomk @grok brows with an icon of some bushy eyebrows

[kaspar @aimuggle]

@sqs @darcydrum @purefunctor @AmpCode was chatting w claude yesterday about their tos because i wanted to roll this myself but decided not to risk it - the compromise might be to run wrap claude code around amp instead?