Plans for the weekend: read the full leaked source code of Claude Code
English
Ronald Rey
1.8K posts
@wesbos So none of this functionality works as expected in other non English languages?
English
@ThePrimeagen Why did this become so popular all the sudden this past few weeks? It's been around for a while but I've seen more content online about it lately specifically
English
@godofprompt The company I work for wouldn't consider even for a second giving AI this much power to our internal technical footprint, and we're nowhere close to the size or impact Amazon has on society.
Feels ironic to me, it should be the other way around
English
This is the most important signal in AI right now, and most people are reading it wrong.
The story isn't "AI broke Amazon."
The story is that the largest cloud infrastructure company on Earth, spending $200 billion on AI this year, still hasn't solved the governance layer between AI-generated code and production systems.
The timeline tells you everything:
→ Amazon mandated 80% weekly AI coding tool adoption
→ Their own Kiro agent was given operator-level permissions with no peer review
→ It autonomously deleted and rebuilt a live AWS environment. 13 hours of downtime.
→ A second AI tool incident followed months later
→ Last Thursday, the retail site went down for 6 hours. 21,000+ users locked out of checkout.
→ Today, a mandatory all-hands to address "a trend of incidents" they can no longer downplay
Amazon's fix: junior and mid-level engineers can no longer push AI-assisted code without senior approval. They're calling it "controlled friction."
That phrase alone should be on every engineering team's wall.
The companies winning with AI coding tools aren't the ones moving fastest.
They're the ones who built review gates, permission boundaries, and deterministic checks before handing agents the keys to production.
Speed without governance isn't velocity. It's liability.
Every team deploying AI in their dev workflow should be asking three questions right now:
What permissions does our AI tooling actually have?
Is there a mandatory human checkpoint before any destructive operation?
Are we tracking AI-assisted changes separately in our deployment pipeline?
Amazon learned this in production. You don't have to.
Lukasz Olejnik@lukOlejnik
Amazon is holding a mandatory meeting about AI breaking its systems. The official framing is "part of normal business." The briefing note describes a trend of incidents with "high blast radius" caused by "Gen-AI assisted changes" for which "best practices and safeguards are not yet fully established." Translation to human language: we gave AI to engineers and things keep breaking? The response for now? Junior and mid-level engineers can no longer push AI-assisted code without a senior signing off. AWS spent 13 hours recovering after its own AI coding tool, asked to make some changes, decided instead to delete and recreate the environment (the software equivalent of fixing a leaky tap by knocking down the wall). Amazon called that an "extremely limited event" (the affected tool served customers in mainland China).
English
Don't get me wrong, AI is a huge productivity boost, but how about we let experienced engineers be the judge and decide when it is optimal to use vs when it is not?
Enforcing and pushing its usage this way is just a recipee for disaster.
English
"their bonuses depend on how much they use AI to write code"
I've seen incentives like this elsewhere, also in implicit form.
Tuki@TukiFromKL
🚨This is so much worse than you think. > Amazon laid off 30,000 engineers. Then told the ones who survived that their bonuses depend on how much they use AI to write code. So engineers started using AI to push changes faster, because their paycheck literally depends on it. > And then the site went down. Multiple times. Amazon's own shopping app broke because AI-generated code got pushed to production. > So what did management do? Did they take responsibility for forcing engineers to use AI they weren't ready for? Did they admit they created the problem? No. They called a mandatory meeting and blamed the engineers. > AI is powerful enough to replace engineers, we've been saying that all day. But it's not powerful enough to replace quality control AND common sense all at once. Amazon proved that executives who don't understand AI are more dangerous than the AI itself. And every company rushing to do the same thing is watching this and learning absolutely nothing.
English
Okay, my journey with @EffectTS_ starts today. Will go through the docs sequentially and start building something incrementally, let's see how it goes.
English
@ipwanciu It's been almost 30 years since outline was released you have to let border go 😂
English
@reyronald Back then, when I started learning dev, outline wasn't a thing yet, so the border stays even nowadays 😑
English
"We're hiring humans" is not a sentence I expected to read/hear in my life time to be honest
English
Final #buildinpublic thoughts on hiring an engineer
Engineer will need to document as they go, like everyone else (including me).
Updates might include:
-> Reproduced the bug locally
-> Found the issue: image resizer not handling SVGs correctly
-> Tried a CSS fix first before touching the component
-> CSS fix didn't work, going deeper into the component
-> Fix working locally, pushing to staging
-> PR open. Tested on Safari, Chrome and Firefox
(these could be automated -> @linear )
Automate Gitbhub to @linear issue status updates
There will be a daily check in on Slack
I am going to remove the Svelte requirement
I am going to keep pay at $9,000 a month and open up the option for a profit share
Continue to share the SOP in the job description
I will keep no recurring meetings but have a one-week onboarding with as many 1:1 calls as required, plus ongoing calls as required
I will try hiring locally in Cape Town and collabing in person
I will do a deeper review after a week and a few weeks
No passive Slack messages - let the person do their work or not, let go of control
English
@helloitsolly The fact that he quit is evidence enough for me to think that he had no ill intent, otherwise he'd have just waited for you to lose your patience and fired him later instead.
Your expectations are not what high-agency engineers in the industry are used to
English
@reyronald They ignored the agreed sop 5 times and didn’t sent the eod update they promised.
English
@helloitsolly As many others have already pointed out, this expectation is not common in the industry and is considered micro-management. Even after reading your listing the nature of the position is not what I'd have expected joining in. This candidate didn't do anything wrong. Best of luck!
English
One of the best websites I've come across recently, I had a bunch of these already that I've saved myself over the years so glad to see them compiled openly in such a neat way
Thank you so much whoever built this
modern-css.com
English
@juliarturc You could do already do this before without skills by somehow having prewritten prompts or context files that you'd manually attach or wire up.
But having this integrated into the model's workflow is a great convenience.
3/3
English
@juliarturc Skills reduce that pain by introducing predictability into the equation. Any mature project will have several tasks associated with it that are recurrent so being able to have precanned instructions for it greatly improves productivity, I'm experiencing this myself now.
2/..
English
I want to understand this so badly, I really do. But why are 1000+ people excited about text files with incomplete documentation...
Am I in the Truman show? I'm not even trying to be reductive, I genuinely want to know what I don't know.
Guillermo Rauch@rauchg
Skills.sh night was packed with 1,000+ RSVPs. We’ll done it again soon!
English
@Bhavani_00007 If you were asking about sharing some secret value with a co-worker for a quick test, you can use a service similar to privatebin.info and have the link expire/burn after reading.
Ideally your company has a self-hosted version of it that's more secure.
English
@Bhavani_00007 Secret managers is the typical way to go as others have shared.
Some solutions exist that allow you to share and commit the .env file openly thanks to encryption, like dotenvx.com
English
how do teams usually share .env variables securely ?
English