Tech News

Z.

Top comment: @Zai_org “GLM-5.2 supports two thinking-effort levels: High and Max. For coding tasks, we recommend using Max effort to enable deeper reasoning and more reliable performance.”

OP

Top comment: @OpenRouter “Notably, the budget panel was comparable with Claude Fable 5 in performance. A panel of Gemini 3 Flash, Kimi K2.6, and DeepSeek V4 Pro, fused together, beat solo GPT-5.5 and solo Opus 4.8 outright. And it landed within 1% of Fable 5 while costing roughly half the price.”

MZ

Top comment: @matei_zaharia “Omnigent focuses on three problems above the level of a single harness: composition, collaboration and control. https://www.databricks.com/blog/introducing-omnigent-meta-harness-combine-control-and-share-your-agents For composition, it lets you create multi-agent teams with different harnesses or swap harness and model mid-session and mid-loop.”

T-

Top comment: @theo “@Teknium Lmao”

GE

Top comment: @GuntherEagleman “@elonmusk Growing by the day!”

TH

Top comment: @tomhfh “@matthewclifford They will in three years time, and everyone will feel completely whacked in the face out of the blue by advancements and dislocations that were entirely predictable.”

NB

Top comment: @nikitabier “@tunguz Is that original content?”

BR

Top comment: @BrianRoemmele “@Scobleizer Robert, loved hanging out last night! I restrained myself from ruining IPO night. These are my thoughts.”

GR

Top comment: @bygregorr “@gokulr @matanSF @FactoryAI @HarryStebbings @20vcFund token hangover is going in the rotation”

J|

Top comment: @RoyalCities “@andersonbcdefg This day is going to be awesome because you'll need to build it again from scratch.”

NO

Top comment: @euxoa “@AndrewCurran_ Maybe not. The RSI in sight will be a relatively thin loop, not a self-maintaining automated system that the government can just watch from the sidelines when it operates.”

MG

Top comment: @matanSF “Full episode: https://youtu.be/lgo_QbgV198”

@L

Top comment: @levelsio “In true Google sense I click the link and don't end up in a flight simulator but in a data layer project manager? @OfficialLoganK”

RS

Top comment: @Scobleizer “@pors Yeah, that was pretty sweet.”

AH

Top comment: @TheAhmadOsman “@MatthewBerman”

AF

Top comment: @badboy999654 “@AshShal @victor207755822 STFU. Anthropic and OpenAI trained their model on the entire internet, pirated books and codebases without consent.”

HH

Top comment: @hopes_revenge “@PeteHegseth @DeptofWar @AnthropicAI you okay buddy ?”

JE

Top comment: @JoshAEngels “My intuition for what’s going on: when we distill from a set of rollouts in SFT, we are establishing a strong prior on a certain assistant persona. Traits of that persona mostly aren’t explicitly optimized against in later training stages like RL and so don’t change that much.”

PC

Top comment: @prz_chojecki “1) Claude Fable‑5‑max is the strongest overall audited public-smoke run. It matches Kimi K2.7 Code on the five accepted solved/settled core results - Problems 1, 3, 4, 5, and 7 - but does so with full 14/14 coverage, the highest average audit score, and broader accepted partial progress. Its advantage is coverage plus depth: it also gives strong usable partials on Problems 2, 8, 9, 10, 11, 195, and 208 without rejected solved claims.”