Everything we publish
Field reports on how these systems actually behave — what was measured, what broke, and what it cost.
GUILD proposes guided data collection loop for imitation learning in robotics
An arXiv paper describes a method that chooses which teleoperated trajectories to collect, rather than filtering a finished dataset.
Paper proposes AI-augmented human intelligence as counterpart to AGI
An arXiv paper argues the field should measure how much intelligence a machine can enable in a person, not just how intelligent a machine can be.
DreamX turns a pretrained video diffusion model into a latent action learner
A new arXiv paper describes a method that compresses video frames into compact latent actions intended to transfer across human and robot embodiments.
Independent re-run of Direct Nash Optimization reports gains on AlpacaEval 2.0
A replication study re-implements DNO-Prct on two open-weight instruct models and reports length-controlled win rate improvements, with some of the gain retained using open-weight annotators.
Ruflo positions itself as an agent harness for swarm deployment on GitHub
A repository called ruflo advertises multi-agent swarms, adaptive memory and federation, but its public description offers no evidence for any of it.
OmniRoute open-sources an AI gateway routing to 375 providers under MIT license
A GitHub project called OmniRoute offers a free MIT-licensed gateway that exposes 1,400-plus models from 375 providers through a single endpoint, according to its repository description.
Why the render died at 3.8 GB
A single ffmpeg filter graph with 84 inputs ate the machine. The fix was thirty lines and a different shape of graph.
Reading the room before writing a line
Marketplaces tell you what they want if you look at what they return for the searches your product should win.
