I know there is a lot of AI slop being submitted here, but I will take the chance.
I built a self-mutating multi-agent coding harness, Varela (named after Francisco Varela, the famous researcher of living systems and intelligence).
Varela has two modes - a nice TUI as well as a Neuromancer-inspired cyberpunk 3D world where you interact with the agents with visual animations of sub agents, tool calls, etc.
The latter is probably not very practical, but I see it as a nice tribute to William Gibson's Neuromancer, which I love. Feel free to submit a cyberdeck as a PR.
I love OpenClaw and what it can do. I have long been wanting OpenClaw to take care of my shopping and purchases online. However, with the current threat of prompt injections, I do not trust it with my credit card details as it stands.
To reduce the chance of breaches, I built ClawLock, a supervisor (sidecar with process isolation) for OpenClaw that injects controls around tool use. The user defines a passport for what the tools can and cannot do and ClawLock governs it. For high-risk transactions like in-browser purchases or shell command, a deterministic policy engine and an LLM-as-judge are used to detect any potential threats prior to execution.
ClawLock is currently in its infancy. It is primarily tested on Linux. I would appreciate any feedback on your experience using it on MacOS and WSL.
I am in strategy consulting and I can tell you the productivity gains are real in terms of research, model building, and summarising work. The result is price pressure from our clients.
This is really cool, nicely done. Q: How did you get the WhatsApp integration working? I thought that was completely closed off unless you are in Europe.
“What this work shows is that the knowledge and reasoning ability of these large language models in an eye health context is now almost indistinguishable from experts,” said Arun Thirunavukarasu, the lead author of a paper on the findings published in PLOS Digital Health journal.”