quality of service

@brendonovich @oofdere What doesn’t work well in WebKit for diffs?

@oofdere as much as mobile support for the web app is nice, it's not a priority atm. we'll support webkit where possible, but some things we use like diffs.com simply can't work as well as they do in chromium

@baleygr728528 @dionisiodev @TestPlace5 This would be like painting something in water color and oil paint, having never done one of them, and then claiming the other superior

@qos @dionisiodev @TestPlace5 if you don't know how to paint or draw, are you not allowed to comment on the beauty, or lack thereof, of the art?

Eu entendo o conceito e as vantagens do Rust, mas, olha que sintaxe feia se comparar com C

@dionisiodev @TestPlace5 Why are you making claims about how ugly a language is if you don’t know how it works?

@TestPlace5 I am not a specialist in Rust 😅

@aidenybai @feulf Average vibe coder. Doesn’t even understand the tools he uses. Just makes stuff to make stuff for no reason

@feulf oh i literally had no idea this existed

introducing Claude Doctor 1. reads your ~/.claude to find where claude keeps messing up 2. writes rules for your CLAUDE.md so it stops. npx claude-doctor

@thdxr @_pi0_ What is fff so good at exactly that ripgrep is not?

in opencode 1.4.4 it no longer spawns or depend on ripgrep we've integrated it natively into opencode thanks to @_pi0_'s wonderful package this is a first step before we move fully over to fff for all our search operations so that things work well on huge codebases

@official_taches Why does this matter? It just dumped them to the terminal

Lol thanks Claude

@aidenybai 1) this is not a skill, this is a CLAUDE.md (memory file) 2) attempting to use any kind of skill or memory file as a safeguard is both idiotic and shows how little you understand about agents

im fully convinced agent skills are just astrology for vibe coders

@neogoose_btw @caidanwilliams Why both?

@caidanwilliams Rust + zig

D̶a̶t̶a̶b̶a̶s̶e̶ ̶+̶ ̶v̶i̶r̶t̶u̶a̶l̶ ̶f̶i̶l̶e̶s̶y̶s̶t̶e̶m̶ ̶a̶b̶s̶t̶r̶a̶c̶t̶i̶o̶n̶ ̶+̶ ̶g̶r̶e̶p̶ ̶i̶s̶ ̶a̶l̶l̶ ̶y̶o̶u̶ ̶n̶e̶e̶d̶ Any filesystem + fff is all you need Efficient file search that automatically suggests recently accessed files, tracks historical data, can automatically expand context, and provide smart fallbacks when your agent is searching for data while keeping extremely high performance bar is what actually you need. Honestly you can call fff a kind of virtual file system optimized for search and automatically syncing with the real one

@elvissun @FastCompany @garrytan Making unmaintainable software isn’t specific to humans. Agents can make it too, and it still results in terrible software that nobody (including agents) can understand or effectively work on

this thread is what mass cope from legacy devs looks like. i talked to @FastCompany about why @garrytan's "AI slop" is actually the future of software engineering. the mass code review. the line-by-line gatekeeping. the "craftsmanship" that was really just slow iteration disguised as rigor - that era is over. and the engineers who built their entire identity around it are panicking. @gregorein brags about burning 3 billion tokens last year while dunking on garry for flexing lines of code. i've burned 6.6 billion in the past three months on codex alone. by his own logic, i'm 8x as credible. see how silly that sounds? yes, he found real issues. yes, they got fixed. that's exactly the point. karpathy's autoresearch proved this already - AI agents can solve very complex problems just by operating inside feedback loops, iterating to optimize a loss function. this is what software engineering is now - gradient descent. ship, measure, self-correct, repeat. all by the agent itself. this is the new startup playbook. your job isn't to review every line before deploy. your job is to build systems where agents observe outcomes - mrr, analytics, error rates, user behavior - and self-improve. the engineer's role shifts from gatekeeper to building the machine that builds the machine. you could run this level of audit (using AI) on any production site and find the same issues - most just don't have a billionaire CEO attached for virality. mocking the people who adapted is easier than adapting. but the craft is evolving whether you like it or not.

@riyandhiman14 @Shpigford @claudeai Claude already has built in policies for that? What???

@Shpigford @claudeai yeah this happened to me sometime back, so i built a small tool that checks every action against a policy before it runs. stuff like git checkout, rm, force push just gets blocked automatically now.

if you're wondering about @claudeai's level of intelligence today. it's not great.

@theo @safirr999 Insane tweet. “Agreed upon value” lmfao! The only value is what people trade the stock at. Same as a prediction market

Stocks are backed by assets with agreed upon value and the market trends upwards over time. If you are a reasonable person, you will absolutely make money on the stock market. Gambling creates no value and is not backed by any real value. It is a race to the bottom. The only winner is the platform hosting it.

This should be illegal

@LLMJunky @NickADobos It’s not really polling. Polling implies the message being sent is not received immediately (push) and rather relies on the consumer to “retrieve” it (pull). However, it’s definitely less good than Claude channels since it requires restarting the watch manually

@qos @NickADobos Another watch cycle sure sounds like polling to me? If the agent has to watch for it that is not what I am talking about

this alone is a reason why mcp will never be "dead" your cli/api cannot ping your active session mid turn.

@LLMJunky A background task notifies Claude (or codex) when it completes. You can use these for notifications, just make a CLI command that blocks and only returns when it wants to notify

@qos How, with what protocol is a cli sending a message to your agent mid turn?

@LLMJunky @NickADobos Background tasks don’t require polling. They return when the notification happens and then the agent starts another watch cycle

@NickADobos thats not 2 way communication though, thats one way communication still (but with extra steps) mcp can ping the agent, doesnt need polling. my question though, is mcp really the final form. maybe there's a new standard we could come up with that makes the best of both worlds

@thdxr Why can’t you just use the API? Anthropic just doesn’t want people using their Claude code subs which is … reasonable

opencode 1.3.0 will no longer autoload the claude max plugin we did our best to convince anthropic to support developer choice but they sent lawyers it's your right to access services however you wish but it is also their right to block whoever they want we can't maintain an official plugin so it's been removed from github and marked deprecated on npm appreciate our partners at openai, github and gitlab who are going the other direction and supporting developer freedom

@ryancarson @rywalker @linear For what reason? What’s wrong with git?

@rywalker @linear A new thing

The company that actually builds the agent-first code factory is going to be worth hundreds of billions of dollars. No one has cracked it yet. It can't be the model labs because then you're tied to one model. I'm hoping a company like @linear will do this. I'd happily pay thousands of dollars a month for that (+ the token cost). Basically, we need SDLC 2.0 for the agent age. (Also, the right solution can't rely on gh - we need whoever does this to completely replace it as well.)

@alxfazio —help exists

mcps may be dead, but no normie is going to install a cli tool + a skill just so the clanker knows how to drive the cli and invoke the skill every time they need to use the tool

@BeardOps @fentasyl Why is that bad? Ripgrep is a better replacement for grep

@fentasyl Codex is obsessed with trying to use ripgrep too instead of just using grep. Odd behavior.

If you ask GPT-5.4-Pro to code for you in Rust, it will spend 30 minutes trying to jailbreak its environment to install the rust toolchain, then give up and use Python as a word processor.

@rafalwilinski @Endlingg CLIs can use oauth… what are you even talking about

@Endlingg Now go tell Karen from your HR department that she needs to use a CLI, generate API key, pull some random skills from this repo and that repo and oh she doesn’t have GitHub account

“MCP is dead” misses the point MCP isn’t about tools. It’s about OAuth If that seems pointless, you haven’t deployed anything beyond a demo