stat-arb portfolio manager pacing the convex hull
Joined October 2021
- Tweets590
- Following265
- Followers92
- Likes594
looking for absolutely cracked people for our new lab effort (think anthropic/evernote/openai caliber)
attention is all you need
data is all you need
compute is all you need
rl is all you need
ya'll we've been bamboozled, seems like we need a lot
this chart makes me very optimistic about open source tbh and maybe the llm paradigm in general. the counterfactual of a narrow open source vs frontier gap would have suggested unfathomable compute and other edge needed to push the frontier even further.
Announcing Artificial Analysis Intelligence Index v5, featuring 2 new benchmarks and 3 upgrades. For the first time, the majority of the Index is built on private datasets
The Artificial Analysis Intelligence Index is our composite measure of model capabilities across tasks reflecting real-world usage today. v5 is our biggest upgrade since v4 launched in January. With v5, 60% of the Index is now based on evaluations with private test sets or held-out answers.
Alongside Intelligence Index v5, weβre launching the Artificial Analysis Progress Index v1, which tracks model capability over time on one continuous scale.
New to the Intelligence Index:
β€ FrontierCode 2.0: Cognition's newly-launched private benchmark of real-world coding tasks, graded on whether a repository maintainer would merge the agent's change
β€ Terminal-Bench-Science: 70 agentic scientific research tasks, run in a terminal environment
Upgraded:
β€ HLE-Diamond (replacing Humanity's Last Exam): a refined 1,000-question version of Humanity's Last Exam, split evenly between reasoning and knowledge questions
β€ AA-Omniscience v1.1: An updated dataset with saturated and public set questions removed, new questions added, and ambiguous or erroneous questions revised
β€ CritPt v2: a private, 40-question hard subset of CritPt v2, revised by the CritPt authors and run with a code-execution tool
Staying in the Index: AA-Briefcase, AutomationBench-AA, GDP.pdf and Terminal-Bench 4.0
Removed from the Index: GDPval-AA v2.1, SciCode, and AA-LCR v1.1
Key results:
β€ Claude Opus 5.5 (max) leads at 58, ahead of GPT-6 Astra (max, 57) and GPT-6.1 Sol (max, 55): Anthropic and OpenAI hold the top five models. Gemini 4 Argon (high) is next at 50
β€ The open weights gap widens from 3 points to 12: Kimi K3.1 Pro (weights not yet released) is the top open weights model at 46. Among released weights, DeepSeek V4.1 Flash 1017 (max) and GLM-5.3 (max) lead at 32 - 26 points behind the leader and below Claude Haiku 5.5 (max, 36)
β€ Cost per task varies by more than 13x among the leading models: of the 10 highest-scoring models shown, GPT-6.1 Terra (max) is the cheapest at $0.96 per Intelligence Index task and Claude Fable 5.1 (max) the most expensive at $13.04
my election prediction is that some people are going to be pretty upset with the outcome (follow for more tips)
against her shoulderβ¨lightly sobbing, fallβs chill airβ¨evening holds us still
(getting liquidated on hyperliquid)
as a relative newcomer to ai itβs cool to fast forward through older discourse and see that everybody was basically right about how ai would turn out a few years later. our current predictions will be pretty accurate for sure.
IF π YOU π HAVE π PURPLE π HAIR π I π DON'T π CARE π WHAT π YOU π HAVE π TO π SAY π I π WILL π FOLLOW π YOU π BLINDLY π TO π THE π END π OF π THE π EARTH
waking up today, lounging in bed
me: isnβt it so weird that weβre married
partner: i ate a full sized twix bar in my dream
malfad as a natural generalization of malfoid
New blog post introducing string diagrams: this is a visual language that we can use to talk about many different kinds of type systems.
Then I use them to show how a monad really *is* a monoid in a category of endofunctors.
roboguy13.github.io/blog/strβ¦
wait so you can just put mts @AnthropicAI or @OpenAI in your profile and then they send pretty people to come flirt with you and talk about math and computers?
dunno how throne works or w/e but i guess if anybody wants to buy me a couple of B300 blackwell ultras that would be kinda cool
Yuhan Fang π³οΈββ§οΈπ¦π³οΈβπ retweeted
"The weakest link in the security chain is the human element" π