The same words can change a vision model’s decision when typography changes
The same words can produce a different judgment when a model sees them as an image. ThinkingType turns presentation drift into a reproducible benchmark.
Archive
AI · page 2
IRZ AI archive, page 2.
Claude is adding invisible marks to its output, but they do not prove who wrote it
Anthropic is starting to mark Claude output with invisible text watermarks and signed provenance for supported files. The useful signal is “processed by Claude,” not “authored by Claude.”
Local AI has a new problem: turning a GPU into a machine people can simply use
Cloudless aims to move beyond DGX Spark onto ordinary GPU computers running Linux and Windows through WSL, without reformatting them. The detail points at local AI’s next bottleneck: making installation disappear.
116 ways an LLM benchmark can be wrong without an obvious error
The community registry tracks 116 serving-path traps where requests look valid, responses look valid, and the final measurement is still wrong because of templates, parsers, kernels, runtimes or version drift.
“AI Engineer” jobs are less ambiguous than the market around them
A video says almost no “AI Engineer” job matches its title. Its source says nearly the opposite: the role is consolidating while the career entry point is getting narrower.
This game agent checks the screen after the code compiles
Godogen turns a prompt into a Godot, Bevy or Babylon.js project. Its most useful detail is not code generation, but a loop that captures frames and repairs visible defects.
Meta fits a 30B local agent model into less than 20 GB
Muse Glimmer is a 30B model built for local agents. Meta says its quantized language model fits under 20 GB and the full inference stack targets a 24 or 32 GB memory envelope.
After 109.8 billion tokens, agents are mostly rereading context
CodeHerder published two months of telemetry from its coding-agent fleet: 109.8 billion tokens. In the priced window, only 0.5% of token volume was output.
An agent memory can become more consistent as it becomes more wrong
Persistent memory can stop an agent from rediscovering everything. It can also preserve a mistake for weeks. Aictx shows why provenance, evidence and staleness need to live inside the memory itself.
At Microsoft, coding agents spread mostly through coworkers
A study of tens of thousands of Microsoft engineers found more merged PRs among CLI-agent users. The more reusable result may be how people started using the tools at all.