AI
Artificial intelligence: LLMs, agents, diffusion, vision, NLP and the latest from top labs.
164 published articles
AI Research
The verification horizon: why verifying coding agents is now harder than building them
A classical intuition holds that verifying a solution is easier than producing one. For today's coding agents, that intuition has inverted: generating complex solutions is now easy. The hard part is reliably verifying them.
2026-07-05
Bot Traffic
Cloudflare to block mixed-use web crawlers by default from September 2026
Cloudflare will block mixed-use crawlers from ad-hosted pages by default from September 2026, targeting bots that blend search, agent, and AI training use. The policy aims to give site owners more control and commercial opportunities as non-human traffic now exceeds human traffic online.
2026-07-05
Deep Learning
M3D and Real-Guidance Bring Dataset Distillation to High-Resolution Realms
Dataset distillation has long been stuck on low-res benchmarks, but a new approach called M3D changes that. By combining multi-scale matching, a data manifold prior, and a Real-guidance strategy, it scales to ImageNet-1K at 128×128 resolution, achieving 68.5% top-1 accuracy with just one image per class and cutting memory usage by ten times.
2026-07-05
Artificial Intelligence
Ai2 cuts satellite imagery AI costs by 3x with a smarter token trick
Ai2's OlmoEarth v1.1 reduces compute costs by up to 3x compared to v1, enabling cheaper large-scale map refreshes. The key innovation is merging resolution-based tokens for Sentinel-2 imagery, cutting token counts by a factor of three while preserving performance through modified pre-training.
2026-07-05
AI Research
How maxproof turns generative verifiers into a proof revolution engine
MaxProof is a test-time scaling framework that models mathematical proof generation as an evolutionary search process. By combining Proof RL, verifier alignment, and refinement augmentation, it turns unreliable generative verification into a trustworthy reward system for training and inference.
2026-07-05
Deep Dive
Aleph Alpha builds theoretical inference model for DeepSeek: Deriving performance from hardware primitives
Aleph Alpha created a theoretical inference model for DeepSeek v3 to estimate throughput from hardware parameters, analyzing trade-offs across GPU setups to help practitioners optimize performance and cost for large MoE models.
2026-07-05
AI Research
Llms corrupt your documents when you delegate: a close look at the delegate-52 benchmark
The DELEGATE-52 benchmark reveals that current LLMs accumulate fidelity degradation when entrusted with multi-step document edits. Errors affect 19–34% of artifact content over 20 iterations, though Python workflows show less than 1% loss. The study is a diagnostic tool, not a verdict on real-world AI utility.
2026-07-05
Model Release
DeepSeek-V4 preview lands, and the open-weight math gets harder for everyone
DeepSeek unveils DeepSeek-V4 preview with top-tier reasoning and stronger agent abilities, now available across platforms. The release signals a major escalation in the open-weight AI race, putting pressure on OpenAI and Google to justify their closed-source strategies.
2026-07-05
Cryptographic Privacy
Vega brings zero-knowledge proofs to identity verification in under 100 ms
Vega enables users to prove facts from government IDs without revealing the credential itself, using ZKPs generated in under 100 ms on commodity hardware. It targets real-world formats like mobile driver's licenses and the EU Digital Identity Wallet, with a fold-and-reuse scheme that reduces proof cost for repeated presentations.
2026-07-04
AI Research & Development
Ma Jiaqi taught MiniMax engineers a hard lesson about forgotten tokens
MiniMax's internal investigation into why its M2 model couldn't output the name 'Ma Jiaqi' revealed a structural mismatch between pre-training vocabulary and post-training data distribution. The root cause: low-frequency tokens' lm_head vectors drift during SFT, losing generation ability while retaining understanding. A full-vocabulary coverage fix resolved the issue and also mitigated language mixing in Japanese.
2026-07-04
Artificial Intelligence
MiniMax's new M2.5 coding model tops the benchmark at 5% of the price
MiniMax's M2.5 model tops the Multi-SWE-Bench coding benchmark, beats mainstream models on workspace tasks, and costs a tenth to a twentieth of competitors. Open-source weights are on HuggingFace.
2026-07-04
Research analysis
AI document corruption in delegated workflows: what a new stress test reveals
The DELEGATE-52 benchmark evaluates AI systems on long-horizon delegated document editing tasks, finding that frontier models accumulate semantic fidelity loss of 19–34% over 20 iterations. Python workflows showed less than 1% degradation on average, but the study underscores that reliable long-horizon delegation remains an open challenge.
2026-07-04