Shorthand comes back
Bringing an early-2000s scripting language back to parity, preparing Apache tests, and getting regular-expression parsing working.
Lynn Cole / Laboratory
Laboratory
Experiments, research notes, mad science, and things in progress.
Shmoosh: a faster packed-attention path
An opt-in TF32 mode makes Shmoosh’s packed-attention path faster while keeping the fidelity policy explicit. Here are the measurements and the remaining tradeoffs.
Shmoosh: controllable attention before dramatic speedups
Shmoosh’s early speed gains are modest, but its attention policies expose useful control over image quality and performance.
Underpaint’s document-scoped prompt memory
Underpaint’s Proompt Manager keeps reusable prompts within the scope of a document, tying the painting workflow to its existing session history.
So how does underpainting work?
Underpainting turns outpainting inward: segmentation supplies the pieces, visual reasoning makes judgments, and diffusion reconstructs what belongs underneath.
Underpaint and a collaborative painting room
Underpaint combines inpainting, outpainting, prompting, detailing and refinement with a plan for collaborative painting sessions.
Long-horizon coding: non-breaking edits and Git
Long-horizon coding needs non-breaking edits and disciplined Git integration, especially when a swarm works across multiple providers.
Fixing the planning loop before the next benchmark run
A new planning loop helped the coding swarm, then conflicting tickets melted the benchmark. The next fix is backlog validation before implementation.
Evaluating a long-horizon solver with Pizza Empire
Pizza Empire tests a long-horizon coding swarm against a whole product: ordering, kitchen workflow, delivery and the constraints that make them coherent.
Build, test, evaluate, audit, refactor
The coding-agent project works on smaller tasks but still fails at its central challenge. Auditing, simplification and another round through the development loop come next.
A coding-agent demo is not yet a product
A coding-agent demo can complete a task. A product needs to understand arbitrary codebases, resolve ambiguity and recover without constant babysitting.
Keeping a coding swarm oriented over long projects
A long-horizon coding swarm needs to stay oriented, inspect itself, recover from mistakes and keep moving through a project that cannot fit in one breath.