Costhanzo_cloro

The plug-and-play design system for AI is progressing, huge references from some are.na projects I found along the way

Ok the typing correction is horrible. iPhone keyboard for multi languages speakers and users is unusable. Why can’t we have a very smaaaaaaall small LLM or model that backward checks the whole phrase instead of word-by-word?

GPT 5.4 is obsessed with calling its subagents "McClintock"

@icanvardar Getting overworked, only scope of Openclaw

still don’t know anyone using openclaw. and still haven’t found a single meaningful use case for it

kokoro-tts for webgpu audio gen in UI is >>>

Mimo v2 Pro on Opencode has replaced Codex for 95% of tasks

@izadoesdev they flipped the chart, both axis

composer 2 is lowkey ass, not sure how it's meant to be better than opus 4.6

I'm overusing QWEN code, it's too fast and too precise compared to any other free tier

@cursor_ai The reason of the inverted graph is to have your model on top right? wtf

Composer 2 is now available in Cursor.

Interfacing

Trying this with @NousResearch's Hermes: added a couple of cron JOBS in a Polymarket bot regulated by a trading profile to enable recursive polishing of results and WR / PNL. Let's see.

Sanity AI agent inside the CMS itself is lifesaving

The work @sanity_io is putting in last changelogs is outstanding

Session: 358 rounds | 43 entered | Win Rate: 100.0% (43W/0L) This @Polymarket thing is getting serious

Session: 250 rounds | 29 entered | Win Rate: 100.0% (29W/0L)

Monochrome UIs are the current most used style - but the issue in having X, Perplexity and Vercel both doing things and not being able to really discern between the three due to the same exact monochrome palette, it's becoming a bit annoying

Session: 560 rounds | 86 entered | Win Rate: 91.9% (79W/7L)

Win Rate: 96.8% (30W/1L)

@perplexity_ai has in the end become my hub for computer-use : emails, text refining, general web search