Arthur Meng (@ArthurCMeng) profile picture

Building @HemeraProtocol | cooking @analystailab | dr @stanford | published @nature prior to GPT era

🌞
Joined March 2020
And in case you wonder why this is an issue, I've been teaching my team, my wife, my friend about "how to actually use Codex" a dozen times 🤣 many of them have been using GPT and thought they were using Codex for months... @thsottiaux
Replying to @thsottiaux
First thing, combine ChatGPT and Codex, at least remove the dropdown menu on the upper left corner of the desktop client app. It will save lots of confusion.
1
1
74
Winning technical benchmarks doesn't mean everybody should trust @claudeai Fable 5.2 with everyday tasks, for example handling my inbox. At @AnalystAIlab we gave it everyday jobs, like a client email and a trip plan. It got many right and slipped on some. nitter.cf/arena/status/212056814…
Big news: Claude Fable 5.2 (Max) by @AnthropicAI takes #1 in Agent Arena with an unprecedented lead: +30.4% net improvement, about 1.6x the #2 model Opus-5.5, across 4.7k real-world agentic sessions from our global community. It's also #1 in Code, Chat and Work. A major leap over Claude Fable 5.1 (Max) - Rank: #5 → #1, Net improvement: +14.8% → +30.4% - Median cost per task: $7.7 → $7.3 (6% lower). The API pricing $10/$50 → $8/$40 per million input/output tokens (20% lower) vs. The Frontiers - #1 Claude Fable 5.1 (Max): +30.4% | $7.3/task - #2 Claude Opus 5.5 (Max):+19.5%) | $5.5/task - #3 Gemini 4 Carbon (High):+17.0%) | $1.5/task - #6 GPT-6-Astra (Max): +14.5% | $6.0/task Breaking down performance across signals: Fable 5.2 improves over 5.1 across all four reported signals, with particularly large gains in user satisfaction and steerability. - #1 Praise vs. Complaint , +73%, biggest jump vs +41% the next model. - #1 Steerability: +33% - #1 Confirmed Success: +22% - #3 Bash Reliability: +14% #1 across all major categories - #1 Code: +34% - #1 Work: +31% - #1 Chat: +27% - Versus Fable 5.1: Code #6 → #1 (+19.8 %), Work #4 → #1 (+15.8 %), Chat #6 → #1 (+12.0%). Huge congrats to the @AnthropicAI team on this incredible release!
1
2
52
On money, it answered a retirement account question using your savings account. It also explained APR, a loan's yearly cost, as if fees weren't part of it. Every assistant in these cases did the same on that one.
1
10
We chose these cases to show where assistants go wrong, so this isn't a full score for Fable 5.2. It answered them later than the other assistants did. Every case, with what each assistant did: analyst.ai/cases/
17
Amazed by how easy it is to identify deep state controlled legacy media : just see if they post about Jenson Huang affairs with Chinese spy, or "must out as CEO".
114
Arthur Meng retweeted
Can you hand your everyday jobs to @claudeai Fable 5.2? You still need to check. Even Claude Fable 5.2 @claudeai got 7 of 17 cases wrong. That includes its own date over your emails' date, a client update you never offered, and a bus and train you'd miss.#AI
2
2
3
13
413
Is it just me, or does Opus feel sharper today than it has all week? Curious if anyone else is noticing it, @claudeai.
1
2
70
Only deepseek seems to be heads down building, instead of playing China AI villain games or scooping mathematicians using their own work.
Your Harness. You Define It. Meet Creator Mode in DeepSeek Harness. Turn ideas into plugins that you can use, from a design workspace to a sandbox game or a 3D LLM visualizer. What will you create?
49
Arthur Meng retweeted
Grok Bot @bot beats Muse for personal matters. We put both through 4 reviews, 450 everyday tasks each. Grok Bot was ahead in all 4, and on trust (96% vs 69%), getting to the point (97% vs 72%), correctness (89% vs 86%) and second tries (90% vs 88%).
6
7
3
16
1,388
Arthur Meng retweeted
We’re testing personal agents on real-life tasks—research, email, travel, personal finance and more—to find out which one you can actually count on. Dots (@ChatGPT), Grok (@grok), Muse (@Meta) and Instinct (@noahrshinn). One made serious errors in both travel and finance: an impossible train connection, then deposit insurance limits from “general knowledge”—no source, no date.👀 Which one was it? Drop your guess below 👇 We’ll reveal it in our next review. #AIAgents
5
4
16
433
Arthur Meng retweeted
This is the future we shall bring into being
32,093
39,929
9,876
330,930
96,502,885
Finally someone speaking the truth...
“0% chance” AI will end humanity by 2030. They must be doing it (the fearmongering) for ulterior reasons. Maybe it's political, maybe it's just attention grabbing." - Jensen Huang's new interview with CBS News, with Jo Ling Kent (@jolingkent ) He pushed back strongly against warnings that increasingly capable AI could escape human control this decade. --- From 'CBS News' YT channel (full video link in comment)
204
真的需要每家都来演戏吗?埋头把模型做好不行吗?
2
115
Arthur Meng retweeted
We were working on the keynote today with @romainhuet and @sama and most of the fun was trying to figure out how to explain it all to you because there is so much good stuff in there that it's a bit ridiculous all in quick succession. We'll have some things next week already to not keep you waiting so long, but very excited to show you all new things we've been working on and how it will all come together in the coming months.
1,067
234
122
8,442
2,071,632
RT @AiRollupai: AI BREAKING 🚨 Anthropic researcher Jacob Coxon has resigned, warning that the race toward superintelligence could become d…
This quoted post is unavailable.
1
2