Matching posts

Matching terms are highlighted

POST

Authors respond on DeepSeek V4.1 Flash CoT controllability: a capability limit, not refusal

The authors say the model tries to follow CoT constraints but fails to execute, so results are a lower bound on controllability, and they plan fine-tuning stress tests on the open weights.

deepseek-v4.1@FangcunLeap　Filed in No.020

1 day agoCopy original

Researcher compares Gemini 4, Astra and Opus 5.5 on paper reading

After reading papers, the user rates Opus 5.5's writing taste, Gemini 4's search strength, and Astra's critical but idiosyncratic lens.

gemini-4@apples_jimmy　Filed in No.008

2 days agoCopy original

POST

Security researcher mixes Hermes Agent, Grok Bot and Muse daily

The author says he opens Grok Bot, Cursor and Muse cloud hosts every day and deliberately spreads work across multiple assistants to avoid lock-in.

personal-agent@evilcos　Filed in No.012

5 days agoCopy original

POST

Researcher used Gemini 4 Argon to draft a paper and mimic his writing style

The author says Gemini 4 Argon wrote the paper's first draft and figures, then rewrote it in his style after being fed his past first-author papers.

gemini-4@jon_barron　Filed in No.008

5 days agoCopy original

Researcher: GPT-6 Astra cracked WWI ciphers via brute force, not insight

A guest says GPT-6 Astra decrypted several unsolved WWI ciphers but only by brute-forcing with the key, not by deriving it, feeding his view that RSI is near.

gpt-6@MTSlive　Filed in No.001

5 days agoCopy original

POST

Researcher criticizes OpenAI's Astra access restriction to security researchers

The poster argues restricting Astra to security researchers is a mistake and proposes vetting 1,000 researchers with heavy usage monitoring instead.

gpt-6@S1r1u5_　Filed in No.001

6 days agoCopy original

Google offers six months of Antigravity Ultra free to students and researchers

Google is giving university students, researchers and postdocs six months of free Antigravity Ultra, including top usage limits, full Gemini 4 Argon access and a science plugin.

gemini-4@EngMoElgaraihy　Filed in No.008

9 days agoCopy original

Ex-DeepMind researchers said to beat GPT-6 Astra on 4 vision benchmarks at 6% cost

A meme-style post claims ex-DeepMind researchers' model beats GPT-6 Astra on four visual benchmarks at just 6% of the cost.

gpt-6@cgtwts　Filed in No.001

12 days agoCopy original

MatrAIx researchers present personal agent evaluation at Google DeepMind

The MatrAIx team shared PersonalAgentEval and Persona-8B in a talk at Google DeepMind, arguing that personalization is key to turning personal assistant models into products.

personal-agent@XiaominLi98　Filed in No.012

11 days agoCopy original

Dutch researcher uses Grok Bot to locate a WWII German radar site

The author fed Grok Bot archive maps, aerial photos and written records from multiple countries to cross-reference, and it successfully located the coordinates of the Klein Heidelberg radar station and its underground bunker.

grok@RobvandenEngel　Filed in No.012

15 days agoCopy original

Showing10of60posts

10 per batch

Keep scrolling to reveal posts 11–20

Showing the first 60 matches. Narrow the query with a model name or tag to find more specific results.