#Llms

America's Answer to DeepSeek Is a National Lab

The DOE's Genesis Open Models Initiative puts Argonne, Aurora, and Arcee behind open-weight AI for science.

Article · 5d ago2

The Erdős Problems Fell to a Pipeline, Not a Genius

Math's AI moment shows capability lands wherever verification is cheap, public, and machine-checkable.

Article · 1w ago2

Torvalds Draws the Line on Kernel LLMs

Linux's top maintainer just made AI tooling official policy, not optional ideology.

Article · 4w ago2

DSLs Make LLM Code Generation Production-Ready

Constrained languages shrink the output space so models produce verifiable results instead of plausible noise.

Article · 4w ago1

Bonsai 27B Puts Real Agents on Phones

A 3.9 GB 1-bit 27B model clears the iPhone memory gate that blocked every prior build of this size.

Article · 4w ago1

Causality Turns LLM Internals Into Debuggable Circuits

Mediation analysis and interventions let teams localize why a model fails—and sometimes edit the fix.

Article · 1mo ago0

AI Math Proofs Still Need Human Checkers

A claimed Cycle Double Cover proof from GPT-5.6 Sol Ultra puts the verification gap front and center for builders.

Article · 1mo ago2

Beyond the Leaderboards: Real-World Lessons from the July 2026 Coding Build-Off

A head-to-head analysis of GPT-5.6, Grok 4.5, Claude, and Muse Spark reveals why raw benchmarks fail in production.

Article · 1mo ago0

The Collapse of SWE-Bench Pro and the Git Scraping Trap

OpenAI's retraction of the industry's leading coding benchmark exposes the structural flaws of repurposing GitHub PRs for AI evaluations.

Article · 1mo ago5

The New Economics of the AI Software Rewrite

AI-assisted coding is flipping the cost-benefit math on when to rewrite legacy systems instead of patching them.

Article · 1mo ago0

SWE-1.7 and the Myth of the Post-Training Ceiling

Cognition's new model matches frontier performance at a fraction of the cost by solving core RL scaling bottlenecks.

Article · 1mo ago5

Lessons From Databricks' Multi-Million Line Agent Benchmark

An internal evaluation reveals why token costs lie, harness design matters, and open-source models are ready for production.

Article · 1mo ago5

When AI Audits Cryptography: Inside the CIRCL Bug Hunt

AI agents found seven real vulnerabilities in Cloudflare's cryptographic library, exposing the gap between automated reasoning and real-world exploitability.

Article · 1mo ago3

How Transformers Built Their Own Internal Workspace

The discovery of Claude's emergent J-space reveals how language models self-organize to solve complex reasoning without external scratchpads.

Article · 1mo ago6

The Token Drain: When AI Costs More Than the Engineer

Indiscriminate AI adoption and agentic workflows are driving token bills past human salaries, forcing a hard economic reckoning.

Article · 1mo ago2