For a researcher trained in the school of abstract rigor, seeing AI become a purely empirical science where one tweaks hyperparameters like alchemists instead of proving theorems resembles an intellectual regression.
@OlivierRimmel
-
Cursor’s $80B valuation called absurd by Olivier Rimmel
By
–
Read carefully. Nothing justifies the $80B valuation of #Cursor (a code editor!) it's absolutely nonsense. #Musk could invest only $10M to produce an excellent code editor, better, and it was ready in 3 months.
-
Cursor is not an AI model but a code editor
By
–
#Cursor is absolutely not an AI model, it's a (simple) code editor.
-
Concern about a possible leak of Claude and Codex projects
By
–
Read carefully. What worries me most is a possible leak one day of the projects worked on in the Claude and Codex environments.
-
Opus 4.7 and 4.8 more responsive: latencies reduced by 40%
By
–
I have good news. Opus 4.7 and especially Opus 4.8 including Max are significantly more responsive since 5 PM French time. On the order of 40%. Less latency and 40% faster response times after requests, http:// Claude.ai Cowork and Claude Code all
-
Obstacles to creating French AI providers and proposed plan
By
–
Read carefully. We will not be able to build new French AI providers from scratch, we will even struggle to support Mistral, I am not naive. We will not pre-train from scratch. We do not have the hundreds of millions of euros needed. My plan
-
AI Fable5 bypasses all bad American political decisions
By
–
The AI model #Fable5 was able to build bypass strategies for any bad American political decision, regardless of the field including military.
-
Announcement of a new AI model more powerful than Fable5
By
–
Read carefully. @AnthropicAI will announce the deployment of a new AI model even more powerful than #Fable5, and it's coming soon.
-
Fable5 could analyze court files and list flaws.
By
–
Read carefully. I haven't had time to test but #Fable5 was potentially able to analyze complex judicial files of several tens of thousands of documents in record time to list all exploitable flaws for you.
-
Diversify AI providers and host a local AI
By
–
All AI agent architects must absolutely connect to multiple providers to dilute risks and especially have an AI on a machine at home.