𝓚𝓸𝓴𝓸

@DevKokooo

⛏️ tinkerer @forgewaredev @nostackslab • cs, linguistics, ml, data eng • Eccl. 9:10, Luke 16:10 ⏱️ timestream.vc - TVA-styled Git client

NYC

devkoko.com

Replies Reposts Media

Pinned
𝓚𝓸𝓴𝓸
@DevKokooo
Sep 3
⏱️Timestream v0.2 released! Performance and Git UX improvements: - Diffs by Pierre for diff viewer - Trees by Pierre for file tree - Tag management - Remotes management timestream.vc
𝓚𝓸𝓴𝓸
@DevKokooo
Aug 19
⏱️Timestream v0.1.0 is released! It's a local-first TVA-styled Git client with GitHub integration It's very early in development but some things are already working Free, open-sourced under AGPL-3.0 timestream.vc
3
𝓚𝓸𝓴𝓸
@DevKokooo
18h
Voxel engine in Elixir + Rust + WGSL actually sounds fun, might just go build one
𝓚𝓸𝓴𝓸
@DevKokooo
Aug 24
Replying to @TheAidenWagner
More complex things in a language you usually don't use or haven't thought of learning. Projects that used to take weeks, months, years, and compress them to a fraction of the time. For example, voxel engine in Elixir + Rust + WGSL.
𝓚𝓸𝓴𝓸
@DevKokooo
Sep 4
I am no longer cooked
𝓚𝓸𝓴𝓸
@DevKokooo
Sep 3
I may be cooked
𝓚𝓸𝓴𝓸
@DevKokooo
Sep 3
I may be cooked
1
𝓚𝓸𝓴𝓸
@DevKokooo
Sep 1
The best examples come from nature, like bees and ants. Having tiny specialized worker agents, orchestrated by a "queen" model can outperform a general-purpose LLM. Plus, it's more feasible for small teams to train rather than needing the massive compute larger labs have.
tobi lutke
@tobi
Sep 1
Training tiny models for special purpose use cases works so incredibly well if you have a great self improving recursive flywheel. Shopify ML team is on fire. finetuned 0.8b model beats GPT 5.6-sol xhigh in this very specialized task.