it’s friday!!!

teaser for this week demo …

if u (like me) no longer google cos claude can do it all in the shell in the flow of coding … shouldn't u care how claude searches the web? does it use google? is google important? r we getting gaslit with search slop?

sunrise …

I've just released the first version of a new Claude Code plugin that helps it remember what it was doing yesterday, why it changed direction, and every conversation it has with you. blog.fsck.com/2025/10/23/e...

@s.ly the writing-skills tends to write skills … shouldnt it try to write (or refactor into) subagent and slash commands too? shouldnt those b other skills as well? i get so frustrated when claude doesnt know how to claude himself

I built a new lightweight Chrome MCP/Skill/Claude Plugin and wrote a bit about the Zen of MCP API Design: blog.fsck.com/2025/10/19/m...

its friday!!!

blog.fsck.com/2025/10/16/s... Anthropic just announced their official skills system across all of Claude. Superpowers is now a Skills plugin.

it’s friday!!

can anybody in github have some decency and add test reports to gha pls … i cannot convince the world that gha is a stupid ci system … could github pls pretty pls try a little bit!!!

i'll take u on this one @pragprog.com just too funny

i just punned the dude ... llmao

todays poster …

tomorrow for vibe coding im gonna take a stab at @emilybache.com lift kata … my spin is with react snaps for approvals and the dude at the wheel … here’s a sample test github.com/limadelic/el... and the schnaps github.com/limadelic/el...

I'm ready to accept a definition of "agent" that I think is widely-enough agreed upon to be useful, at least in the AI engineering space: An LLM agent runs tools in a loop to achieve a goal (I've been dismissing the term as hopelessly ambiguous for years) simonwillison.net/2025/Sep/18/...

i dont say enuf how much i hate TS

almost in a ben afflect way lmao

and i thought adam carolla was the charismatic one lmao

isn’t surprising that 2 hosts of the man show had become so prominent… i confess i watched that show lmao … girls jumping on trampolines lmao … i think im wiser now i hope lmao

upgraded to tahoe … made me feel old .. do i need glasses to see the glass?

unpopular idea? clean code > tested code

watching another coding agent demo where the first question is what is this project about … so after the agent filled ur context with that great question … what can u do? how can we make this question obsolete?

what im doing is having claude and codex bidding over my next diff … yeah im very zen … so instead of trying to multitask im trying to get more quality faster … i pair program im not PR driven

i know yall trying to get multiple things in parallel with agents … ive been however trying to do 1 thing with multiple agents instead … i noticed @kentbeck.com is experimenting with it … any more takers out there? tidyfirst.substack.com/p/genie-figh...

ended up with codex side by side claude each one blind bidding for the next change :) … i keep telling the loser what the other won with lmao

claude broke for an 1 or so switched to codex … humm it weird it felt efficient but hard to understand … idk .. idk … like i coukd not talk to it n riddles a lot … but felt better in some way

so tired of this mystical prompt whispering … theres no magic ppl … if u communicate clearly with humans u can prompt

ur PR should not b legacy code b4 hitting main … that’s not the definition of legacy

what really pisses me off is that ai is way better at cleaning than at generating

and yes ur code is effing dirt … im not gonna comment on the gazzillion thing u could have done with ai to keep it clean !!! PR approved!!! code SHIT SHIT SHIT

RANT!! will we ever effing use ai to cleanup code !!! dam ppl … that’s why crack was an epidemic … yall chasing the next hit instead of getting high on cleaning ur own supply

just cos ur paranoid does not mean the ai is not lying to u

A small rant on the contemporary Zeitgeist for the weekend. Fuck 996. lucumr.pocoo.org/2025/9/4/996/

It's finally happening!

getting close to vibe cleaning ... i can already get the dude to apply my styles until clean ... and i already can see clean file that i dont know exactly what they do but they look like very clean

introducing Speck

i feel i live in a perennial state of gaslighted by the dude

im thinking a thread tool ... the agent would decide which context is a thread and would put in context based on the question ... is that concept out there?

the dude likes my tests too much lmao

im starting to use claude code in a diff way ... i open multiple terms on same dir no worktree stuff ... the idea is i use diff threads of context on same folder ... like deep thinking in one ... vibe coding another ... refactoring/cleaning in other

thinking on dci for agents as a tool

ok my agent recommends me consistently as the best candidate for my position ... my job here is done ... lmao

the one thing i learn on GPT 5 launch ... chat is cat in french!! hence

hey im not the only one saying it!!

programming 3.0 kickstarted the day the llm was blackboxed ... that day signaled the resurgence of the devs ... datascience belongs in the box ... lmao

the test output