Tech investor. A skeptical empiricist surrendering to life; Studying fat tails, removing blind spots, seeking dynamic quality, and letting go. Views are my own.
New York, NY
Joined June 2010
- Tweets48.3K
- Following7.5K
- Followers2.8K
- Likes195K
Dickson Pau retweeted
Day 25/
We are doubling all usage limits for all Pro 100/200/500 tiers and Business Premium, effective today.
We heard your kind feedback and won’t be reducing usage limits for Pro 200, so your Work and Codex usage will stay the same. We’re also going to continue offering extended limits for dots for now, and usage of your dot will still continue to be separate from your default usage.
And we'll of course also keep shipping better models and other efficiency improvements that increase your effective usage over time.
Wow
Day 25/
We are doubling all usage limits for all Pro 100/200/500 tiers and Business Premium, effective today.
We heard your kind feedback and won’t be reducing usage limits for Pro 200, so your Work and Codex usage will stay the same. We’re also going to continue offering extended limits for dots for now, and usage of your dot will still continue to be separate from your default usage.
And we'll of course also keep shipping better models and other efficiency improvements that increase your effective usage over time.
Dickson Pau retweeted
It's so hard to internalize that models get better at an increasing pace.
I talk to plenty of people at Anthropic, OpenAI, etc who have a hard time wrapping their heads around it. Even Claude struggles.
I asked Opus 3 when it thinks I'll have PRs written entirely by LLMs. It guessed no sooner than 2028. Opus 5.5 expected ETH after 2045.
Little baby Claude has no idea how quickly it's growing up and I think it's even harder for humans.
It's finally here!!
Free newsletter: Anthropic’s S-1 uses an anomalous quarter to con investors into becoming its exit liquidity, and explicitly states that its costs will grow aggressively as margins decline. Its $430bn in non-cancellable contracts are yet to start hitting.
wheresyoured.at/exit-liquidi…
Dickson Pau retweeted
Meet NVIDIA Nemotron 3.5 Transcribe 🎙️
An open-weight, promptable 30B MoE model with ~4B active parameters that brings transcription in 40 languages, speaker attribution, translation, contextual prompting and audio understanding into one checkpoint.
Watch it keep track of who said what when everyone’s talking at once.
1/ 🧵
Dickson Pau retweeted
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, we’re launching the Artificial Analysis Progress Index v1, which tracks model capability over time on one continuous scale.
New to the Intelligence Index:
➤ 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
➤ Terminal-Bench-Science: 70 agentic scientific research tasks, run in a terminal environment
Upgraded:
➤ 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
➤ AA-Omniscience v1.1: An updated dataset with saturated and public set questions removed, new questions added, and ambiguous or erroneous questions revised
➤ 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:
➤ 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
➤ 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)
➤ 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
Okay this now looks much more reasonable
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, we’re launching the Artificial Analysis Progress Index v1, which tracks model capability over time on one continuous scale.
New to the Intelligence Index:
➤ 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
➤ Terminal-Bench-Science: 70 agentic scientific research tasks, run in a terminal environment
Upgraded:
➤ 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
➤ AA-Omniscience v1.1: An updated dataset with saturated and public set questions removed, new questions added, and ambiguous or erroneous questions revised
➤ 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:
➤ 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
➤ 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)
➤ 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
Dickson Pau retweeted
Codex's Dreaming V3 memory system uses GPT-5.6 Luna and Terra by default.
Upgrade by copying the following into config.toml:
extract_model = "gpt-6-luna"
consolidation_model = "gpt-6.1-terra"
Both models are more intelligent AND cheaper than their predecessors.
It's better than a free lunch. It's like getting paid to eat lunch.
Bro, many of these points seem too reasonable to come from him
Zitron will tweet some mix of:
Decline rate on models being very high. Useful life of like ~6 months agains big training spend.
Revenue gains being token pricing which is undermined by OS models.
Commitment sizes against competitive uncertainty/lack of visibility into the future.
Ant said margins might go down so today's picture is unsustainable.
Fraud (if he's desperate)
48 😭😭😭
OpenAI has released GPT-6.1 Terra, pushing the cost efficiency frontier. It is the first Terra-class model since GPT-5.6, and is more intelligent for less than a third of the cost
Pricing per token is half or less that of GPT-5.6 Terra - at $1/$5 per million input/output tokens vs $2/$12 for the previous model. Cache write has also halved, now at $1.25 per million tokens. Cached input is one quarter at $0.05 vs $0.20 per million tokens.
Key takeaways:
➤ Pareto-optimal for Intelligence vs Cost at low and medium effort: At max effort, GPT-6.1 Terra gains 6 points in the Artificial Analysis Intelligence Index over GPT-5.6 Terra, at less than a third of the cost per task ($0.40 vs $1.40). At lower effort levels the gain is larger, between 7 and 13 points. The low and medium effort levels are Pareto-optimal for cost efficiency, while the other effort levels sit behind the frontier. At max effort, the model scores 4 points below GPT-6.1 Sol and 10 points above GPT-6 Luna in the Intelligence Index.
➤ Pushes the Coding Agent Index cost efficiency frontier: With a score of 48 in the newly released Artificial Analysis Coding Agent Index v2.0, the model is 6 points below GPT-6.1 Sol (54) and 17 points above GPT-6 Luna (31). The xhigh and low effort settings sit on the frontier of cost efficiency in the Coding Agent Index, while other effort levels are behind the frontier.
➤ 50 point drop in hallucination rate: GPT-6.1 Terra guesses 38% of the time when it does not know the answer to a question in AA-Omniscience, our private knowledge and hallucination benchmark. Previously, GPT-5.6 Terra's hallucination rate was 88%. This significant reduction in hallucination is coupled with a 2 point increase in AA-Omniscience accuracy - meaning the model simultaneously answers more questions correctly.
➤ 20 point jump in Terminal-Bench 4.0, small improvements across most evaluations: The model makes significant progress in the terminal-based agentic benchmark, Terminal-Bench 4.0, with a gain of ~20 points, now rivaling GPT-6.1 Sol at max effort. Gains across most other evaluations are smaller, for example, 5 points in GDP.pdf.
Congratulations @OpenAI and @sama on the launch!
No one folds their foldable 75 times per day
just 75,000 folds is embarassing as hell.
for context that roughly means:
75 folds/day: 3 years
100 folds/day: 2 years
150 folds/day: 1.4 years
we still don’t have real independent tests on the Z Fold8, but Samsung rates it for 500,000 folds.
that means it can last 13 years with 100 folds/days, which is 11 years longer than the Duo.
gap vs. the Duo is crazy even if you give it some margin.
Basically trash
NVIDIA Nemotron 3.5 Super VL from @NVIDIAAI on ARC-AGI (Verified):
- ARC-AGI-2: 1.94%, $0.02/task
- ARC-AGI-1: 22.67%, $0.03/task
Nemotron 3.5 Super VL is the first NVIDIA model we've evaluated on ARC-AGI.
