Bienvenue sur mon blog

Reflexions sur la technologie, le developpement et l'innovation

Articles récents

3 Juin

Microsoft's MAI-Thinking-1 goes toe to toe with Claude Sonnet 4.6

Microsoft dropped its first reasoning model at Build 2026 Day 2, and the benchmarks are worth paying attention to. MAI-Thinking-1, a 35-billion active parameter Mixture of Experts model (roughly 1 trillion total parameters), claims human preference parity with Anthropic’s Claude Sonnet 4.6 in blind evaluations. It also matches Claude Opus 4.6 on SWE-Bench Pro. That is a lot of firepower for a model that costs less to run than anything in Opus or Sonnet’s weight class. The key claims from the Mic

Lire l'article
2 Juin

Microsoft builds its own coding AI and cuts the OpenAI cord

Microsoft Build 2026 opened today in San Francisco and the headline was not Windows. Satya Nadella unveiled Project Polaris, a homegrown AI coding model that will replace GPT-4 Turbo as the default brain inside GitHub Copilot starting August 2026. Automatic migration for all Copilot subscribers, with a three-month fallback window for teams not ready to switch. This is not a minor model swap. Polaris runs on Microsoft’s custom Maia 200 AI accelerators inside Azure, which the company says cuts per

Lire l'article
1 Juin

First confirmed LLM agent cyberattack exfiltrated AWS database in under an hour

Sysdig documented something the security industry has been warning about but had not yet seen in the wild: a large language model agent autonomously performing post-exploitation in a real attack. The target was an AWS-hosted PostgreSQL database. The entire chain, from initial access to exfiltration, took a little over an hour. The attack happened on May 10, 2026. The entry point was CVE-2026-39987, a pre-authentication remote code execution vulnerability in Marimo notebooks. An unauthenticated a

Lire l'article
31 Mai

Nightmare Eclipse released six Windows zero-days in six weeks before GitHub and GitLab pulled the plug

A pseudonymous security researcher operating under the name Nightmare Eclipse spent six weeks systematically dropping working Windows zero-day exploits on GitHub. Six exploits. Six weeks. All targeting different parts of the Windows security architecture. Two code-hosting platforms eventually banned the account. The exploits are still out there. The campaign started April 3 with BlueHammer, a Windows Defender local privilege escalation flaw. The PoC gave attackers a SYSTEM shell via a TOCTOU and

Lire l'article
30 Mai

AI found 10000 critical bugs and the patches still are not keeping up

Anthropic published the first progress report on Project Glasswing on May 22. The numbers are staggering. In roughly 30 days, Claude Mythos Preview and about 50 partner organizations flagged 23,019 potential vulnerabilities in open-source software. Of those, 6,202 were classified as high or critical severity across more than 1,000 projects. Independent validation confirmed 1,726 as true positives. 1,094 earned a high or critical rating. Only 97 have been patched. That gap between discovery and r

Lire l'article
29 Mai

Anthropic ships Opus 4.8 and a $65 billion round on the same day

Anthropic released Claude Opus 4.8 and closed a $65 billion Series H at a $965 billion valuation within hours of each other. The timing was not accidental. The new model landed 41 days after Opus 4.7. That is unusually fast for an Anthropic flagship. The most recent Sonnet and Haiku models are three and seven months old respectively. The accelerated cycle was likely driven by the lukewarm reception Opus 4.7 got from developers, combined with competitive pressure from OpenAI’s Codex and Google’s

Lire l'article

Mots-clés

$40 billion $65 billion 1m context 2.8t parameters 3d AI Agents AI efficiency Application Loader Cross-Platform Cybersecurity DKIM DMARC DeepSeek Development Encryption GPT-5 HTTPS LLM LLM optimization Latency Medusa framework Network Protocols Nuitka Open Source PyInstaller Python Reinforcement Learning Rust SPF Software Distribution Sparse Attention TLS 1.3 Web Performance academic integrity acquisition agentic ai agentic ransomware agents ai ai acquisition ai agents ai alignment ai apps ai chips ai code review ai coding ai compute ai costs ai cybersecurity ai economics ai encyclical ai ethics ai infrastructure ai investment ai kill switch ai layoffs ai model ai models ai office ai overview ai partnership ai pricing ai regulation ai safety ai search ai security ai slop ai spending ai super app ai threat defense ai training data ai ultra ai valuation akamai alexander hanff alibaba alphafold amazon amazon bedrock android anodot anthropic anti-spoofing antitrust anysphere apache 2.0 app store appfigures apple arch linux artificial analysis arxiv ascend ascend 910c ashley macisaac assured robot intelligence async aur authentication bypass automatic translation autoregressive decoding aws aws exfiltration azure badhost benchmarks biometrics biotech bitlocker bypass blackmail botnet bug bounty build 2026 bun caisi canvas cbrs cerebras character.ai chatgpt check point china china ai chips christopher olah cisa cisco class action claude claude code claude fable 5 claude mythos claude opus 5 claude sonnet 4.6 cloud cloud computing cloudflare code security codex coding agent coding agents coding model colorado ai act colossus compute congress consent containment copyright coreweave covert networks cpanel credential theft critical infrastructure csam curl cursor cve cve-2026-35616 cve-2026-39987 cve-2026-41940 cve-2026-42208 cyberattack cybersecurity cybersecurity vendor data breach data center data exfiltration data leak daybreak deepfake deepfake ban deepseek deepseek v4 deepseek-v4 defamation developer tools diffusion language model digital services act digital world conference diplomat directory enumeration distillation distributed inference dma drug discovery dspark duckduckgo dutch court easy-day-js ebpf rootkit ed zitron edge computing edtech education election security elon musk email security ems enterprise ai enterprise browser eu eu ai act european commission exploitarium export control export controls extortion fable 5 fair use financials fingerprints finland fireeye firewall credentials foreign intelligence fortibleed forticlient fortinet four-day-workweek frontier ai fugu gabbard gdpr gemini gemini 3.5 flash gemini nano gemini omni gemini spark gemma 4 generative ai geneva geoffrey hinton gitar gitea github github actions github ban github breach github copilot glm 5.2 godfather of ai google google chrome google deepmind google gemini google i/o 2026 google io 2026 google play government stake gpt-5.5 gpt-5.6 gpt-rosalind gpu capacity gpu pooling gpus in space great american ai act grok grok 4.5 gta 6 hallucination happy oyster hardware healthcare high-risk ai huawei hudson rock hugging face humanoid robots i-dlm industrial-policy inference inference acceleration inference optimization infostealer infrastructure instructure investment ios ipo iroh itron jadepuffer kepler communications kimi k3 lakeview langflow lapd large language model large language models lawsuit layerx layoffs lee jae myung libssh2 life sciences linux malware litellm lithuania llm llm agent llm deception llm inference lockdown mode longcat losses luna magnifica humanitas mai-thinking-1 maia 200 mandiant mastra mcafee enterprise meituan memory mesh llm meta microsoft midterms mini shai-hulud mixture of experts mobile apps model routing moonshot ai mortgage multi-agent systems multimodal musk mythos mythos 5 nasdaq national registry national security ncsc nebius nightmare eclipse noam shazeer non-consensual imagery north korea npm nvidia nvidia gb300 nx console nyc oci odni oidc omnibus open source open source ai open weight open weights openai openai trial opus 4.8 oracle oracle cloud orbital compute orchestration p2p parallel decoding parallel token validation patching bottleneck peer preservation peer review performance phishing detection photo editing piracy police records policy pope leo xiv post-training pre-deployment testing privacy private cloud compute project glasswing project perception project polaris prompt injection public wealth fund pulte pwn2own qwen3.8 raas ransomware reasoning model research robot-tax robotics rockstar games russia rust s-1 sakana ai salesforce sam altman samsung sandbox escape search data search engine sec filing security self-hosting semiconductor semiconductors series h settlement shinyhunters silent safeguards silicon valley siri ai sk hynix smart meters snowflake software development software engineering sol sol ultra sora source code breach south korea space technology spacex spacexai spcx speculative decoding sql injection stargate state preemption state-sponsored hacking student data subagents superapp superintelligence superintelligence labs supply chain supply chain attack swe-bench sysdig tang tan tanstack teampcp tech jobs tech layoffs terminal-bench terra the gentlemen third-party vendor threat intelligence together ai token pricing tokio tpu trade secrets training data trainium transformer tree-based attention trellix trivy trump trump ai order uber uc berkeley us-china user backlash vibe coding voting machines vpn vs code extension vulnerabilities vulnerability vulnerability detection vulnerability disclosure vulnerability reporting wafer-scale engine web hosting web security white house windows 11 windows zero-day wiper wiz workforce automation world model worldleaks wwdc 2026 x xai xdr z.ai zero trust zero-day zig zuckerberg