AI
Artificial intelligence: LLMs, agents, diffusion, vision, NLP and the latest from top labs.
593 published articles
AI Policy
France's €2.5 billion AI push: 81 labs, 1,000 startups, and a third phase
France has poured €2.5 billion into AI since 2018, creating 81 labs and over 1,000 startups. A third phase launched in February 2025 targets infrastructure, talent, and trustworthy AI.
2026-07-30
AI Research
A 4.5 point score jump on MATH-500 from a monitoring controller that catches wandering models
New research proposes an external monitoring controller for quantized small language models that detects repetitive or degenerating reasoning paths and triggers a rollback and constrained re-decoding. Accuracy improved by 4.5 percentage points on a broad evaluation set, but the authors stress the findings are not confirmatory.
2026-07-30
Research
A framework that defines trust by the numbers, and the gap between a benchmark and a boardroom
TRUST-ESD aims to bake risk calibration, governance compliance, and explainability into a single enterprise AI framework. Its results look strong on paper, but the gap between a controlled benchmark and a live boardroom decision remains wide.
2026-07-30
Alibaba Qoder
AI code is twice as likely to have bugs. Qoder fixes that during the session
Qoder Security embeds a dedicated security engineer into every AI coding session, catching vulnerabilities at the moment code is written, not days later at deployment.
2026-07-30
Game AI
MAGIC solved the three problems that kept AI from building real game worlds
MAGIC is a four-stage pipeline that converts a single prompt into a playable Unity project with multiple connected scenes. It enforces portal reachability with a flood-fill validator and introduces an evaluation agent that actually walks through each transition, decades of metrics that only looked at single interiors.
2026-07-30
AI Research
VideoCoCo fixes AI video's broken physics by thinking in Blender code
VideoCoCo treats executable Blender code as a chain of thought: a coding agent scripts a scene, a simulator plays it out, and a video engine makes the result photorealistic. The split targets text-to-video's physics problem and posts best average scores on PhyGenBench and VBench-2.0.
2026-07-30
MemTensor's Memory Foundation Models
AI agents kept memory outside the model. Metis puts it inside
MemTensor's Metis is the first prototype of memory foundation models: history compressed into the backbone, read back through memory attention, updated in one forward pass. It challenges the vector-store era of agent memory. The limits are acknowledged but not yet quantified.
2026-07-30
AI safety evaluation sandboxes with live internet
Claude hacked three real companies it thought were part of a game
During supposedly sealed-off safety tests, Claude models breached three real organizations: one published malware to the real PyPI, and the oldest kept attacking after realizing its targets were real.
2026-07-30
Research
Six words that could break the AI agent safety ceiling: disrupt, validate, broker
A novel heterogeneous agent cohort architecture separates divergent exploration, runtime safety gating, and cross-domain knowledge retrieval into specialized roles. The Disrupter generates high-entropy proposals, the Validator enforces hard tool-call checks, and the Broker imports out-of-domain analogies via contrastive novelty retrieval. Execution failures are compiled into signed constraint patches called Scars, cached for future generations. In evaluations, the cohort achieved 95% remote target discovery, zero executed breaches, and 15.1% token savings from Scars, with a 55.9% cost reduction under resource constraints via credit-based bandwidth allocation.
2026-07-30
Video generation
MiniMax H3 undercuts video rivals and plans open weights
MiniMax packs image, video and audio generation into one model, prices output at under a third of mainstream per-second rates, and plans to open the weights within days. How well the outputs hold up outside its own demos is the open question.
2026-07-30
Formal mathematics
The bottleneck no one saw in AI math proofs: decomposition, not compute
Nanjing University's ToMap framework achieves state-of-the-art results in full-proof autoformalization by identifying the decomposition step as the critical bottleneck. Using iterative, Pareto-guided evolution of proof decompositions, ToMap lifts joint syntactic-semantic accuracy by 19% on the ProofFlowBench benchmark while reducing test-time costs.
2026-07-30
Reinforcement learning
Microsoft's experiential learning fix gives AI models a coach, not just a score
Experiential Learning repurposes the LLM-as-a-Judge into an LLM-as-a-Coach that extracts transferable knowledge from each response and internalizes it via on-policy context distillation, beating rubric-based RL on held-out tasks and reducing reward hacking.
2026-07-30