here's the part that stuck with me. claude outperformed every human candidate who's ever taken anthropic's internal engineering hiring test. not most. every single one in the company's history. a 2-hour rigorous assessment designed to filter for elite engineers. so what
MACHINE LEARNING
-
Claude Opus 4.5 hits 80.9% on SWE-bench, first to break 80%
By
–
the benchmark data made me pause. claude opus 4.5 hit 80.9% on SWE-bench verified. first model to ever break 80%. this isn't leetcode. it's real github issues from production repos. the actual work developers do. 4 out of 5 real-world bugs. solved.
-
When does AI assistance become replacement? 60% engineering work on Claude
By
–
i looked up the internal numbers. their engineers are using claude for 60% of their work. not "sometimes helps with debugging." sixty percent of everything. 50% productivity gains. 2-3x better than a year ago.
at what point do we stop calling this "assistance"? -
Daniela Amodei says Claude matches human-level coding at Anthropic
By
–
daniela amodei on CNBC: "by some definitions of that, we've already surpassed" human-level AI. her example was coding. she said claude writes code "about as well as many developers at anthropic now." anthropic. the company that probably employs some of the best engineers on
-
Top AI Research Papers: RAG, Agents, and Memory Optimization
By
–
The Top AI Papers of the Week (January 12-18) – UniversalRAG
– Agent-as-a-Judge
– Self-Evolving Search Agents
– Active Context Compression
– Efficient Lifelong Memory for LLM Agents
– Extending Context by Dropping Positional Embeddings
– Unified Long-Term and Short-Term Memory -
AI Accelerates Scientific Progress Solving Unsolvable Math Problems
By
–
Nous observons les premiers signes d’une accélération ×1000 du progrès scientifique et technologique, portée par l’IA. L’IA résout désormais des problèmes mathématiques extrêmement difficiles (presque chaque jour) que les humains n’avaient jamais réussi à résoudre jusqu’ici. Ce
-
ScaledML Companies Achieve Massive Valuation Growth and Success
By
–
The alpha in ScaledML companies is insane: Groq, Cerebras, OpenAI, Tesla, Google, Meta, Databricks, Matroid and many others have all at least 10x'ed in valuation since first presenting at ScaledML.
→ View original post on X — @reza_zadeh, 2026-01-17 21:21 UTC
-
LLMs Replace Hand-Crafted Rewards in Multi-Agent RL
By
–
A huge claim from this paper on the end of reward engineering. Reward engineering remains a persistent bottleneck in multi-agent RL. This paper argues that LLMs enable a fundamental shift: from hand-crafted reward functions to natural language objectives. If language can
-
Robot companies building AI infrastructure beyond video models for sim-to-real transfer
By
–
Oh, believe me, the robot companies aren't only doing video models to run the robots. There's a whole AI infrastructure to build the software that gets loaded into the robots. You're absolutely right that they need to worry about the sim-to-real transfer and making sure that
-
Chinese AI dominates video models: 7 of 8 top, faster, cheaper
By
–
But here's what nobody's saying: Chinese AI just dominated the leaderboard.
7 of 8 top video models: China. Faster. Cheaper. Better. While everyone waited for Sora, they shipped the future.