Mino (@compounded_mind) header
Mino (@compounded_mind) profile picture

AI news, research and bad takes. accelerationist. professional shitposter in training per aspera ad astra

Joined March 2024
“few weeks” 🙂
Replying to @ishuagra02
Yeah but now it’s a few weeks
1
115
OpenAI is giving four external research organizations deeper access to its internal models and systems to investigate alignment failures throughout development, not just before release Apollo Research will investigate whether models can secretly work against their developers' intentions and whether existing safeguards can detect it METR will examine misaligned behavior during reinforcement learning, including failures that current monitors may have missed Transluce will audit post-training alignment evaluations, develop additional tests and identify gaps in their coverage Goodfire will use interpretability tools to analyze model internals, helping researchers investigate training behavior, debug failures and improve alignment
We’re working with @ApolloResearch, @METR_Evals, @TransluceAI, and @GoodfireAI on embedded assessments: giving outside experts deep access to our internal models and systems to investigate alignment and test our safeguards. These assessments build on our ongoing alignment and safety research, bringing independent scrutiny to how we identify and manage risks in our systems. openai.com/index/our-first-p…
2
131
the US Commerce Department just announced a new rule targeting remote access to advanced American AI chips through data centers in third countries
1
1
103
the cost comparison makes this even less encouraging $3.27 per task for Qwen4-Max versus $1.27 for GPT-6.1 Sol on Artificial Analysis Alibaba reports major efficiency improvements over Qwen3.8-Max, but that doesn’t necessarily make it competitive with the alternatives
1
35
Qwen4-Max looked pretty promising from Alibaba’s benchmark charts, but the independent results are rather disappointing 43 on the Artificial Analysis Intelligence Index, with a reported cost per task roughly 2.6x higher than GPT-6.1 Sol despite scoring significantly lower really hoping the smaller open-weight models turn out better
📢 Meet Qwen4-Max, your new AI teammate for work, creativity, and discovery! Qwen4-Max is built on a new MoE architecture with native vision and a 1M-token context window. Highlights: - Outperforms Qwen3.8-Max with up to ~50% fewer tokens, while average cost per task drops over 60% in our internal evaluation. - Advances real-world productivity through greater proficiency with professional tools, stronger long-horizon execution, and higher-quality deliverables. - Handles real tasks, including 3D modeling, game development, manufacturing, finance, law, healthcare, and frontier scientific discovery. - $1.5/M tokens for input, $4.5/M tokens for output, and $0.075/M tokens for implicit caching — all lower than Qwen3.8-Max, with implicit caching down by 70%. The API is live on QwenCloud. You can also try Qwen4-Max on Qwen Studio, QwenWork and Qoder. Start building today! 🚀 - Blog: qwen.ai/blog?id=qwen4-max - API: qwencloud.com/models/qwen4-m… - Token Plan: qwencloud.com/pricing/token-… More to come! The open weights of smaller Qwen4 models will be released. Stay tuned. 👀
1
1
139
even @patience_cave is roasting OpenAI now 😭😭
250
just read the replies under this post OpenAI announces a genuinely good improvement to Codex, something people have been complaining about for ages, and half the comments are about how everyone has moved to Claude i mean, what did you expect? people are frustrated, they’ve been waiting for a new model, and every extra week only makes that frustration worse. i’ve already moved a bunch of my own tasks to Gemini, and i doubt i’m the only one. at this point, even genuinely good news gets buried under all that frustration really hoping we finally get something this week
Codex team might have actually fixed the "model at capacity" errors that have been pestering you.
9
22
6,060
Jensen Huang is reportedly considering a return as NVIDIA CEO, just days after his resignation the FT reports he’s discussed the possibility with confidants, while an ally says they’re assembling a Silicon Valley “war room” to lobby NVIDIA’s board and seek White House support, claiming Jensen himself backs the effort if true, this is no longer just a public campaign to bring him back. Jensen may be actively supporting it himself
1
41
OpenAI’s first science blog post details GPT-6 Astra’s role in a candidate solution to a long-standing problem in particle physics Nathaniel Craig supplied the original strategy. Astra suggested the refinement construction, proved convergence and led the detailed mathematical argument three outside lattice theorists also examined the work and challenged aspects of the proof as it developed Edward Witten has since published a separate note discussing Craig’s construction and a related approach
Physicist Nathaniel Craig shares how he worked with GPT-6 Astra to develop a candidate lattice definition of the Standard Model. It’s the first post on our new science blog: openai.com/index/step-toward…
9
1,735