faisal (@faisal_sayed05) header
faisal (@faisal_sayed05) profile picture

waterloo se | ai/robotics/bio @medra_ai prev. @thoughtfullab @wisprflow @primer @whop @heynox @hackclub

sf
Joined February 2021
Photo by faisal (@faisal_sayed05)Photo by faisal (@faisal_sayed05)Photo by faisal (@faisal_sayed05)Photo by faisal (@faisal_sayed05)Photo by faisal (@faisal_sayed05)Photo by faisal (@faisal_sayed05)Photo by faisal (@faisal_sayed05)Photo by faisal (@faisal_sayed05)Photo by faisal (@faisal_sayed05)Photo by faisal (@faisal_sayed05)
thinky put diligence bench on their model card :)
Introducing Inkling 2 Small. Scoring 31 on the AA Intelligence Index, it's a big jump over its predecessor. Inkling 2 Small is available for fine-tuning and inference on Tinker today, with full weight release planned soon. thinkingmachines.ai/news/ink…
1
8
661
where’s the line wtf
1
10
628
one man’s inshallah is another man’s mashallah
1
2
25
727
watch this space
Who grades the graders? LLM-as-judge on short answers is well studied. Agent judges are not: they open the files, run the code, and investigate before ruling on another agent's work. Introducing Arbiter-Bench: 71 real agent runs that frontier agent judges get wrong.
3
693
glad to know that the biggest ai lab in the world also faces the same issues with markdown rendering iykyk
3
569
somehow i am more excited to build something with jev than with gpt 6 astra
4
254
i know i’m a little late but i’m actually so confused by all these fly demos did they not simply just release a mapping of all the neurons of a fly and no understanding of how these neurons fire or what the response to stimuli looks like are these demos just claude making stuff up and gaslighting people that they’re actually controlling a fly???
1
12
620
does anyone know where i can get/borrow ~50 disposable Ag/AgCl electrodes in SF for free
1
340
Today, @paperinstr and @thoughtfullab are releasing DiligenceBench, an agent-first benchmark for long-form equity research. DiligenceBench is designed to be a simple, grounded playground for our rubric-based RL and harness optimization work.
8
12
1
48
4,091
the chud yearns for nacho fries
161
4,423
269
64,120
928,906
why do all these car dealerships in the bay area have an ai assistant answering your calls which yc company did this
3
23
3,337
Opus 4.8 outperforms every other model on AttuneBench - best at picking the response humans actually preferred - biggest MSCEIT four-branch jump of any Opus generation - entire pairwise top-4 is now Anthropic models. non-Anthropic frontiers stall ~50%
Introducing Claude Opus 4.8: it builds on Opus 4.7 with sharper judgment, more honesty about its own progress, and the ability to work independently for longer than its predecessors. Available today at the same price.
5
44
8,067
faisal retweeted
In post-training, we've learned that once a behavior is measurable, you can train AI to excel at it. EQ is one of the hardest things to verify. AttuneBench makes it measurable through observable signals: whether a model notices distress, tracks shifting preferences, adapts to context, and responds in a way people experience as helpful.
Introducing AttuneBench! We built this benchmark on a simple premise: for self-improving AI to reach its full usefulness to humanity, it needs high EQ. We decomposed EQ into distinct skills and evaluated 11 frontier models across 50 real-life topics, from relationships and marriage to school and job stress, using 50,000+ first-person annotations.
6
9
3
80
13,287
reminder that ~22% of 18–21 year olds in the US have used LLMs for mental health advice and the best frontier model scored 54% on our EQ benchmark it has never been more important to be thoughtful join us
Introducing AttuneBench! We built this benchmark on a simple premise: for self-improving AI to reach its full usefulness to humanity, it needs high EQ. We decomposed EQ into distinct skills and evaluated 11 frontier models across 50 real-life topics, from relationships and marriage to school and job stress, using 50,000+ first-person annotations.
1
8
1,050
1/ Today we're releasing AttuneBench, the first open EQ benchmark grounded in real multi-turn human-model conversations, scored against what the person actually felt and wanted at each turn. Built by the research team at @pareto_ai in collaboration with @thoughtfullab. Most existing EQ benchmarks rely on: - synthetic prompts - single-turn interactions - third-party annotation None directly measure how a model reads and responds to a real person across a full conversation. We evaluated 11 leading models from major providers, across 200 conversations and 50,000+ first-person annotations.
14
26
5
151
23,432
We tested 11 frontier LLMs on 200 real human–AI conversations to measure emotional intelligence The result that surprised us: EQ doesn't scale with size or recency. Claude Haiku 4.5 beats Sonnet 4.6. Opus 4.6 performs better than 4.7 It's an orthogonal capability and labs aren't optimizing for it
9
10
5
115
11,803
i made ai agents play among us 10 frontier models locked in a room, forced to lie, deceive, and gaslight each other to make it out alive
3
1
1
19
2,693
black is innocent here. he still got voted out lol watching these agents talk to each other and decide on what to do is kinda fun play / spectate: amongus.fayd.me/ source: github.com/faisalsayed10/amo…
1
361