2026-10-29

Diffusion video models produce striking clips, but they cannot reliably render an exact formula, a correct chart, a rule-consistent game or narration synchronized to on-screen events. A different route is emerging: multimodal agents that write code to render every frame and synthesize every sound. We introduce VibeVideoBench, a benchmark for evaluating agents on code-driven video generation. Its 239 self-contained tasks in English and Chinese span 12 categories, from data stories and music-drive

🤖VibeVideoBench evaluates multimodal agents that generate videos via code, finding that top systems meet core requirements on most tasks but frequently fail due to poor time management.

Multimodal intelligence can be viewed as human-guided co-evolution between worlds and agents through interaction. However, scaling this process faces two bottlenecks: constructing heterogeneous, verifiable environments is costly, and manual Harness iteration cannot keep pace with growing task diversity and complexity. A key insight is that tasks with different modalities and surface forms can share the same underlying capability requirements. We therefore introduce The Trinity, a recursive self-

🤖The Trinity framework recursively improves multimodal agents by organizing tasks into a capability tree, generating simulated training worlds, and routing specialized harness modules to boost performance on diverse benchmarks.

Memory is increasingly used in long-horizon agents, yet the net value of building memory over direct retrieval from raw history remains largely unclear. Through a large-scale evaluation across multiple benchmarks and model backbones, we find many dedicated memory systems underperform naive RAG in many cases, indicating a negative net value. Taking RAG as a reference, we dissect the net value of memory into the gains and losses over direct retrieval, termed memory benefit and memory tax. These ga

🤖Large-scale evaluation reveals that dedicated memory systems often underperform naive retrieval-augmented generation, introducing a framework of memory benefit and memory tax to assess net value.

Recent test-time training (TTT) frameworks have enabled open-weight large language models (LLMs) to outperform inference-only frameworks powered by frontier models on scientific discovery tasks. However, TTT introduces additional computational cost for online weight updates during the discovery process, making computational efficiency an important consideration when scaling to larger models. This motivates us to explore an efficient test-time adaptation method for scientific discovery. We introd

🤖Test-Time Memory Evolution updates external memory without changing model weights to guide scientific discovery more efficiently than test-time training.

Training a spatial VLM to answer which side an object lies on does not require it to preserve the metric displacement that determines that answer, while supervision for a distance estimate may leave categorical relations largely unconstrained. When post-training emphasizes one readout, improvements can therefore concentrate on that interface while transfer to others remains weak or even degrades; we call this observed specialization interface lock-in. ChronoReplay addresses it by replaying heter

🤖ChronoReplay mitigates interface lock-in in spatial vision-language models by replaying heterogeneous spatial supervision chronologically, improving cross-task transfer and enabling efficient inference-time evidence selection.

I think October represents another inflection point in capabilities for coding agents and the models that drive them: Claude Opus 5.5 can now build video games that are actually fun simonwillison.net/2026/Oct/28/...

A hacker obtained personally identifiable information about more than 40,000 Homeland Security staffers, which has since been published online.

Tech leaders are flaunting how out of touch they are when they brag about using AI to connect with their families.

But the AI responses get more forthright when asked outside the US.

The MPC Sample Key isn't even formally announced yet, but Akai is already showing off what the modding community is cooking up through vibe coding. Akai has decided to embrace the explosion of hackers and custom firmware developers by giving them early access to the device, which will have open-source firmware. It will likely run […]

I think a turn of events are pointing towards a very likely crash in prices in the coming months. With a single gb300 rack being to serve all of Openrouter's traffic for DSV4.1 Flash( ) Half of data center buildouts already being cancelled/stalled. Neoclouds unable to finance debt for their compute, and being stuck in a 5 year deal with no confirmed demand. AI adoption not matching expectations. Smaller models being able to serve majority of the requests. OpenAI/Anthropic being unlikely to finan