I create Neurosymbolic AI using first Oder logic and predicate calculus alongside our own agentic AI stack and LLMs. @pierrepinna
LLMS
-
Sparse Attention and Large Context Windows: An Industry Shift in AI Models
By
–
确实很酷，拿 2900 万美金做 12M context，
— 艾略特 (@elliotchen100) 5 mai 2026
侧面证明了一件事：整个行业都开始相信稀疏注意力是 dense attention 的解药。
SubQ 走的是「重训一个模型」，属于垂直整合，风险大回报也大。@evermind 的 MSA 走的是「给主流模型加记忆」，属于 水平嵌入，谁的模型都能用。
另外，SubQ API 跟 SubQ… https://t.co/yZ8IugqquAYeah, that's pretty cool—using $29 million to build a 12M context window, which indirectly proves one thing: the entire industry is starting to believe that sparse attention is the antidote to dense attention. SubQ takes the "retrain a single model" approach, which is vertical
-
Builders Share Feedback on GPT-5.5 After Weeks of Testing
By
–
Excited for 5:55 tonight! Wish we had room for everyone. We’ll host more builder events soon. Also curious what you all think of GPT-5.5 after building with it for a couple weeks!
-
GPT-5.5 Boosts Contributions Across All OpenAI Teams
By
–
And @AaronFriel, who works on engineering acceleration at @OpenAI.
— Romain Huet (@romainhuet) 5 mai 2026
From his view, GPT-5.5 is changing far more than backend work:
“We’re not just seeing an increase in contributions from engineers and researchers, but from everyone across the company.” pic.twitter.com/jdOY6bsD0pAnd @AaronFriel
, who works on engineering acceleration at @OpenAI
. From his view, GPT-5.5 is changing far more than backend work: “We’re not just seeing an increase in contributions from engineers and researchers, but from everyone across the company.” -
GPT-5.5 Resolves 98% of Bugs Autonomously in Real Workflow
By
–
Today is 5/5! GPT-5.5 shipped 12 days ago, and the community love has been amazing. ✨
— Romain Huet (@romainhuet) 5 mai 2026
Revisiting a few first impressions we hadn’t posted here before seeing more of you tonight at 5:55.
First: @clairevo gave GPT-5.5 a backlog of bugs, and it did “a 98% job all by itself.” pic.twitter.com/ZdOL8QGUiiToday is 5/5! GPT-5.5 shipped 12 days ago, and the community love has been amazing. Revisiting a few first impressions we hadn’t posted here before seeing more of you tonight at 5:55. First: @clairevo gave GPT-5.5 a backlog of bugs, and it did “a 98% job all by itself.”
-
GPT-5.5 Achieves Record Financial Document Extraction at Ramp
By
–
Next: @willkoh_kc from @tryramp tested GPT-5.5 inside their own harness.
— Romain Huet (@romainhuet) 5 mai 2026
“It was discovering ways to use the tools that we had given it… and figuring out novel ways to solve problems.”
He also tested it on financial-doc evals: GPT-5.5 hit their highest perfect extraction rate. pic.twitter.com/4C1N8QIrscNext: @willkoh_kc from @tryramp tested GPT-5.5 inside their own harness. “It was discovering ways to use the tools that we had given it… and figuring out novel ways to solve problems.” He also tested it on financial-doc evals: GPT-5.5 hit their highest perfect extraction rate.
-
HuggingFace bf16 Positional Embeddings Bug Went Unnoticed
By
–
Remember how HF precalculated positional embeddings in bf16 for a long time and the impact was small enough that it slipped through unnoticed? :O
-
Positional Embeddings Were Less Important Than Originally Thought
By
–
It actually turned out the particular spectrum of the chosen waves often is pretty horrible, but it also turned out positional embeddings weren't actually that important anyway so no-one really noticed for years.
-
X Post Quality Tip: Character Limits for AI-Generated Content
By
–
We wanted the singularity; we got the circularity.