lack

@_kelindi until LLM code gen came along: webxos.itch.io/vaultbreaker

every software engineer is just someone who wanted to make video games and quietly gave up on that dream

@0xSero kimi and deepseek are very close in UI design, kimi has made some exceptional work for its capabilities with html/ui/css work

I've gone back to using Kimi-K2.6 for almost everything UI related. It's just so much better than anything on the market rn. Opus-4.7 seems to have the same high-IQ related visual degeneracy you see in GPT-5.5 models. Sonnet is too sloppy to trust on my codebases. Kimi 3 wen

@elliotarledge LACK: under development - pretty much baking in every core function a local agent harness needs. Limited to ollama models for now for testing, but the progress has been great so far it works very well even with contrained small models. Embed model moderation is my current idea:

this post has some interesting replies. what are YOUR thoughts on local model harnesses?

Embedding model is used because it's lightweight, fast, and sufficient for the task (no text generation needed). The moderator acts purely through post-processing + rule-based logic + linting, not through LLM generation. It injects strong feedback into conversation context.

Turning @ollama nomic-embed-text model into a moderator for LACK was a success. Will be out in a new update. LINT code review, git pushing and auto moderation with a pure embed model:

github.com/webxos/lack Under Development

AX explained!

Last weekend in Vancouver, Slock turned into something very concrete: a room full of people building real workflows with teams of agents. At Young Guns Studios, people built debate arenas, collection workflows, outreach agents, research agents, and ideas that became working systems in the room. Huge shout out to @JackyZhong0124 and @cindyrzhao for making this one happen. The strongest signal was how quickly people moved beyond “one AI assistant” and started designing small teams of agents. That shift — from tool to team — is what we’ll keep building toward.

@lack2026 @ollama The embed model approach sidesteps the latency trap of running full LLMs for moderation. Most teams default to bigger models when smaller ones trained on semantic distance actually catch the nuance without the overhead.

Starship flip and landing burn at the end of its twelfth flight test

Raptor 3 may turn out to be the most important thing created by humans if it leads to us becoming multiplanetary. This demonstration makes it feel inevitable. This is my most detailed shot of the engines in flight. I added these prints to my shop, linked in the reply

LACK v3.6.4 is now live: github.com/webxos/lack Real-time Graph Fully Perfected, /graph Command Enhanced, SLIME Mobile Interface Upgraded, Code Quality & Maintainability, Reliability & Safety Fixes + more.