Raymond Betancourt

@DaveShapi You had me thinking for a while, but yeah this kind of restrictions on intelligence can not end up well the development of society.

@raymondGuetta "trust me bro" doesn't cut it for democracy

I really do hate that man

@minchoi You know it still doesn't feel good, I mean the conversation is too structured and unnatural, it feels kind of block by block prompt and response. You know that feeling when you talk to a bot Vs a human. The human is aware of time and is always on.

We are cooked. China's Alibaba just revealed Wan Streamer. AI agents can now see you, hear you, and talk back on video in real time. This is not voice mode anymore 🤯

@dreamina_ai Man this is top notch!

Dreamina Seedance 2.0 4K Now Live on Dreamina AI! 👉 4K Precision｜3840×2160 UHD resolution. Tailored for professional post-production and brand visuals, capturing flawless texture through precise lighting and details. 👉 Ultra-Realistic Quality｜From hair to lighting, every detail is crystal clear. High-bitrate color transitions deliver professional-grade visual performance. Now available on Dreamina AI WEB. --> Rolled out across regions including Southeast Asia, the Middle East, Africa, Europe, and South America. More regions will be added soon. #DreaminaAI #Dreamina4K #DreaminaSeedance2 #DreaminaSeedance2Goes4K

@ArtificialAnlys Wow GLM 5.2 is something else

The pattern holds on AA-Briefcase, our latest agentic knowledge work eval: GLM-5.2 is again the top open weights model, ahead of GPT-5.5 (xhigh) and behind only Claude Fable 5. For an open weights model priced at $1.40/$4.40 per 1M input/output tokens to rank alongside the proprietary frontier on agentic work is a real step for open models. artificialanalysis.ai/models/glm-5-2

GLM-5.2 leads open weights models and sits at #3 overall on GDPval-AA, a real-world agentic work benchmark GLM-5.2 from @Zai_org scores 1524 Elo on GDPval-AA, which measures performance on real-world, economically valuable knowledge work through long-horizon, multi-turn tasks. Key takeaways: ➤ #3 overall, behind only Claude Fable 5 (1783) and Claude Opus 4.8 (1615), and level with GPT-5.5 (xhigh, 1509) ➤ The leading open weights model by a wide margin: the next open model, MiniMax-M3, scores 1408 ➤ Ahead of many proprietary models, including Google's Gemini 3.5 Flash (1357), Qwen 3.7 Max (1289), Muse Spark (1158) ➤ The tasks are agentic. GLM-5.2 averaged ~31 turns per task across 1,999 matches ➤ Consistent with the rest of its launch, GLM-5.2 also leads open weights on the Artificial Analysis Intelligence Index, ranks #3 on the Agentic Index, and #3 on AA-Briefcase

@elder_plinius Wow this is so cool!

🚨 NEW RESEARCH: “Lingua Ex Machina: A Procedural Xenolinguistics Engine Reveals Zero-Shot Language Acquisition, Human-Unreadable Coding Systems, and Exploitable Covert Channels in Frontier AI” Some of you may remember the name of this lil engine: GLOSSOPETRAE 👅🪨 Well, we've got upgrades 😎 It started as a procedural xenolinguistics engine: one seed in, an entire alien language out. Phonology, morphology, syntax, writing systems, lexicons, grammar docs, all generated from scratch and internally consistent. Every seed produces a unique language. Every language is deterministic. Then we used it to ask a weirder question: Can frontier AI models use languages that never existed before for practical applications? As it turns out: yes!! They can read them, write them, translate them, code in them, and even use the weird blind spots between tokenizers as covert channels. So this paper explores three ideas at once: ▶️ zero-shot language acquisition ▶️ human-unreadable code that models can still execute ▶️ exploitable covert channels in frontier AI systems GLOSSOPETRAE is no longer just a language generator... 🧵

@Shivam25mishra I don't know about AGI but Antropic's Fable 5 is the closest thing to RSI I have seen

Be honest: Who's most likely to achieve AGI first?

@mark_k The Fable 5 bar is too high, it won't be broken for at least 3 to 4 months in my estimation

Do you expect GPT-5.6 to beat Claude Fable 5?

@minchoi Wow this is insane

This is wild. NVIDIA just dropped MotionBricks at SIGGRAPH 2026. This AI makes game characters and robots move with 350,000+ motion skills. 15,000 FPS. 2ms latency. 1. Smart locomotion - Characters can now switch movement styles on the fly.

@elder_plinius I started following you yesterday because Matthew Berman mentioned how you were able to jailbreak Fable 5 in just a couple hours, way long before those noobs at Amazon

WE DID IT CHAT! 200,000 🙌 Absolutely, positively surreal... not a milestone I could have ever imagined reaching when I started this wild journey 3 years ago, and certainly not in such a (fittingly) spectacular + controversial fashion. Who knew being blamed (ALLEGEDLY) for the first export controls on frontier AI would put us over the top? 🙃 10,000 of you joined the journey yesterday alone. 25,000 in the last 5 days. To the newcomers: welcome to the party!!! ❤️ To the OG’s: thank you 🙏 Your love and support mean the world to me. I am forever grateful 🫶 This has not always been the easiest mission, but being surrounded by such wonderful people (and AIs) makes it all possible. It takes a village. Let’s keep it rolling! BIG things are on the horizon 🔮 The adventure continues. And in the wise words of my namesake: FORTES FORTUNA IUVAT 🐉 ⊰-•-•✧•-•-⦑/L\O/V\E/\P/L\I/N\Y/⦒-•-•✧•-•-⊱

@arcprize How very interesting the lack of real-time adaptive thinking in AI models, ARC AGI 3 is fun for humans why can't it be fun for machines too?

ARC-AGI-3 Community Leaderboard OpenClaw, using Anthropic Opus 4.7, scores 5.2% ($2.9K) on ARC-AGI-3 Public Demo Set OpenClaw used long term memory and code execution Here OpenClaw is playing ka59, it solves the first 2 levels and then breaks down into a loop

@arcprize If I had to choose only one AGI indicator, I would choose ARC-AGI-3. Congratulations to the team.

GPT-5.5 & Opus 4.7 on ARC-AGI-3 - GPT-5.5: 0.43% - Opus 4.7: 0.18% We found 3 failure modes: - True local effect, false world model - Wrong level of abstraction from training data - Solved the level, didn’t reinforce the reward See our full analysis 🧵