ryan

@nobodyshomeuk @mitchellh oh I get the point, and it's oblivious to the fact that humans have written more slop than AI ever will.

@RyanPCooke @mitchellh His argument doesn't boil down to "humans are superior". Like he's said twice, you're missing the point. He wants AI PR's that are Human reviewed and actually TESTED. Not AI slop or garbage prompted PR's that wastes him and other contributors time. You just fail to see that.

silly humans who still think they are better at coding than datacenters full of PHDs, resorting to sabotage

@archidapp i'm 100% a silly human

@RyanPCooke What's wrong with having human reviewers. You don't realize the irony of calling humans silly, when you can only view the argument in terms of 2 extremes?

@masylum ha!

@RyanPCooke LLMs are not better than Mitchell

2025: AI can write code, but not in production use cases 2026: AI can write production code, but humans still need to review PRs can't wait for 2027

@mitchellh your argument unavoidably boils down to, in some scenarios, humans are superior to ai at coding. even if that scenario is catching a dumb human trick. i get it, it's hard to fully embrace the idea that ai is better than you at your job. i struggle too big fan of your work btw

@RyanPCooke I'll let you think about it.

@mitchellh how does bad human drivers evaluating their own driving filter out anything?

@RyanPCooke You're missing it. Garbage in garbage out. Bad AI drivers produce bad AI results. That's not an AI problem, its a people problem. I'm extremely, EXTREMELY PRO AI (see my blog, projects, I embed helpful AGENTS.md, skills, etc.). This filters out bad AI drivers.

@mitchellh but it is the point. if you truly believe the machines are better, you should be incentivizing ai prs, not humiliating them

@RyanPCooke That’s not the point. The poisoning catches stupid PEOPLE. The machines are fine.

LinkedIn messaging is embarrassingly bad. So I built a faster, private, open-source inbox. ⌨️ Superhuman-style shortcuts ⚡ Instant search with local message sync 🌙 Dark mode, emoji, and more 🔒 Runs completely local as a Chrome extension Open source and free. Get it below 👇

@signulll half agree. I love waymo, super fun to ride in, but I never use a car in NYC. rational take is that good public transit killed it

what a fucking tragedy for nyc. the same dumb resistance happened during the uber saga & now you’re watching the same shit play out again with autonomous cars. the reality is that humans should not be operating metal death machines forever. esp in a city where taxis have spent decades gaming routes, ignoring pickups, driving like maniacs, & treating the meter like a side quest. uber fixed some of that, then slowly became expensive as hell, degraded, & unreliable in its own stupid way. collective decision making of humans is often so so damn stupid.

@RhysSullivan yup, made my first presentation in HTML a few weeks ago. so much more slick than anything I could have cooked up with slides

underrated aspect of the ai wave is never having to deal with a .pptx or .docx file again html and markdown have won

@ciruai thanks for posting this. claude initially told me that the NPU wasn't going to be useful for me

My local LLM rig is run headlessly in my closet on an AMD Strix Halo. It streams games and movies to my house, runs 24/7 hermes, manages home services. It all works even when the internet goes out. I even use the NPU for uninterrupted always on inference in addition to the iGPU with very little additional overhead. My wife and son interact with it without knowing how it works. People say the AMD system is too slow or not useful for real work. I strongly disagree. There is definitely a knowledge gap but there are a lot of people working to help bridge it and it's not as hard as people think. In this video I exceed 110 token/s with qwen3.6 35b pretty consistently, all while the NPU is chugging along on it's own little model. I use web chat in a nice ui, hermes and pi over the tui But before you think "oh thats all technical terminal stuff" you can point this api endpoint to any application you want. The only difference between this and ollama is about 200-500% more power with the same models. Things are only going to get better.

LAUNCH WEEK is next week 🍿 incredible work by the @workos web team 🔥 (see if you can find the easter eggs 👀)

I've been trying to think of the best analogy for model prompt routing. it's not tcp/ip because there I still specify the address I want to reach (domain name). or maybe it is because I should be prompting for goals not directives?

@cassidoo we've started doing "Claude days" where we pair an engineer and non-engineer and the mandate is to build an app that's been on the back of their minds

Honestly I love hackathons, BUT I do think: - internally many orgs run them as "fix the backlog" time - externally they're often like "pls give us API use case ideas" When they're actually fun, I *love* them. The latest ones I've done have been game jams for that reason.

introducing the greptile partner program. we worked with top dev tool companies like stripe, openai, and datadog to teach greptile how best to use their APIs when greptile detects that your code touches a partner’s API, it reviews the code with additional context from the partner. it then flags the issue along with a link to the docs

@JunaidAckroyd hate to burst your bubble but AI has been writing better code than any dev for the past 6 months

Devs, what’s your plan B if AI writes better code than you in the next 6 months?

@engineering_bae biased, WorkOS

We’ve heard a ton about companies who have a trash engineering culture. What companies still have a solid engineering culture despite our tech world being so in flux? Obviously no company or team is perfect, I’m looking for ones doing more good than harm.

@verycracked Brooklyn if you want to feel like you are living in a TV show, Queens if you want to feel like you are living in NYC

Alright nyc folks, let’s say hypothetically I move to nyc, should I be looking at Queens or Brooklyn? We’re a family of three and a large dog.

@daniel_tian1 cool, congrats on the launch!

We're launching the world's 1st iMessage App API for agents. your agents can now send Apps inside iMessage - an Uber Eats menu, a mini game, or any agentic UI. no other app, no website. welcome to the iMessage Era try it now: miniapp.photon.codes