Hacker Newsnew | past | comments | ask | show | jobs | submit | askonomm's commentslogin

I'm good, thanks. AI bros can all gather at Omarchy if they want slopOS experience.

Seeing the sheer level of incompetence in the U.S government, I don't think citing CIA has the intended effect you think it has.

CIA in 2000’s wasn’t yet under trump government and even then, they are still one of the most secure organizations.

Trump or not, everyone think of them as highly secure.


You mean the same organization that leaked chats by using Signal and had people in the group chat that wasn't supposed to be there?

Finally doing something standards-compliant instead of forcing users into a proprietary workflow.

fwiw, i am with thariq https://x.com/trq212/status/2092302273099796842 in that prompts should be tuned for models and in fact blindly applying agents.md is probably an antipattern unless you want all models to basically converge to some common ill defined of instruction following - good local minima, bad global minima for model diversity and exploration of intelligence.

aka, sometimes it really is too early to force a standard


Is it realistic to rewrite your AGENTS.md every six weeks? That's about how often Anthropic releases a new point release of Opus.

You tell Opus to do it.

I have had very little luck with agents.md. What has worked well for me is a ./docs folder. They seem to just create and update stuff on their own.

If you want this it's trivial to add an AGENTS.md that simply says "if you're Claude read CLAUDE.md, if you're Astra read ASTRA.md". A common entry point is good regardless.

This wastes both tokens and turns. But yes it's probably the best option we have today.

It can try its own file and fall back to generic like here. What's wrong with that?

Wasting turns? That is silly, use a better harness. Also token usage can mitigated by incremental discovery instead of stuck 5k+ worth of tokens in the AGENT/Claude md file.

Every turn means more tokens in ways that are not obvious to most people and lead to tons of unnecessary cache reads.

No harness can batch your agents.md read with the reads the contents of the file tell it to read.


That's of course rather nonsensical.

In a "one LLM only" environment, your instructions are by default tuned for said LLM.

In a multi-LLM environment, roughly nobody will keep separate sets of instructions for each. It's not a realistic take.

On top of that: If your LLM is so bad at reading that it can't follow a set of instructions that wasn't specifically written just for that one single precious LLM, I sure wonder what that says about your employers repeated statements that ASI is definitely right around the corner.


By this logic you'd probably be wise to tier your claude.md by model (sonnet/opus) as well as effort level too, considering the varying failure modes

except they have similar pretrain/rlhf data which is the thing u really want to tune for

YMMV but for me even models in the same family fail in different ways, and every incremental update changes it

depends what you're doing. if you've got a specialized agent deployed in prod, of course your evals and prompts will be targeted towards 1 specific version of a model.

on the other hand if it's just a local coding/"use my computer" agent, i highly doubt the effort in maintaining different prompts is worth any gain in performance


19 out of 20 harnesses supporting the standard isn't "too early".

Tariq is wrong and it's not an antipattern. Reason being that a good AGENTS.md impacts all models in a positive manner. If it affects certain models negatively, it means you're putting the wrong things in it.


No thanks, I'm not tuning a bunch of files just for things to break when I switch models or a new one comes out.

I'll just use my one-size-fits-all AGENTS.md file and tweak it when the one of the clankers screw up. I don't have time for such busywork.

Actually, I will append extra rules to CLAUDE.md (which imports AGENTS.md) since there is a hook there, and Claude has its own foibles. So I'll backpedal a bit there.


It looks for Claude.md first so I don't understand what you think the problem is with the standard name as a fallback.

Yeah but are models good enough to review these files and say “i would work better if you worded it this way?”

CLAUDE.md: @AGENTS.md

    ln -s AGENTS.md CLAUDE.md

Wake up babe, new recursive self learning technique just dropped.

Why not hardlink - so it wouldn't even know?

Because Git can track symlinks and not hard links since they look like ordinary files.

I tried that before, it doesn’t work. Claude will not prioritize AGENTS instructions the same way it did for CLAUDE.

ln was the only thing that worked for me


Trying to make a .md file proprietary by changing the name is hilarious

claude.md predates agents.md.

Yeah but .md predates claude.md. I had agent.md files in my repo before Claude tried to act like it's a special protocol.

The world has moved on.

Standards? In LLM harnesses? I wish

I'm honestly not sure if this is tongue in cheek and the "finally" is in the silly way it is often used, but the claude.md variant existed first. Indeed, the agents.md thing was pretty clearly a "that's neat, let's do that with a different name".

It's "finally" because people have been asking for it for a long time. No one cares that "claude was first," what they want is for Ant to follow the conventions and not put extra work on us. This was such a minimal thing to do, and considering how much they vibe and claim "coding is solved," we thought it would not be too difficult to respect AGENTS.md, so finally seeing it happen, while nice, is too late for me. I've moved on from Big Ai and only use open weight models now.

Just because something is first, doesn't mean it's the standard. The standard is put in place by how many harnesses use a specific pattern, and that has been AGENTS.md for a long while now. Long enough that Claude Code not supporting it is really just an act of forcing people into a proprietary workflow, because of course it's in their interest to make moving away as painful as possible given how little moat the LLM models themselves have. In fact, I think Claude Code has been pretty much the only one not supporting AGENTS.md. Check out https://agents.md if you want more information on it.

There’s also that whole MCP thing.

They did start with C++. They experimented with Swift, then experimented with Rust, ultimately choosing Rust. I don't see how experimenting with things makes anything "not serious". If you don't question your primitives and just accept them "because that's how it has always been", you're not a very good engineer in my book. In any case, the tech stack is not in constant flux as you claim, it's still C++, with more and more parts moving / being started in Rust, with the eventual probability of Rust replacing C++ organically, I suppose. Seems like pretty sound strategy to me to slowly, but surely, adopt a memory-safe language.

I don't know about UK, but everything in the U.S constitution's first amendment is most definitely also in Estonia. I would pretty confidently say also most of EU, probably. They are pretty basic things.

Why is it that a lot of the rich and powerful people tend to become so childish? Or is it just more visible when rich and powerful people do that? Either way, it’s just cringe throwing tantrums like this.

What I’d love to see is the ability to use my own domain name, so that if that service ever goes under, I can set up a git service with the same paths and everyone relying on my software can continue to do so. It’s just hard to find the trust to invest in a completely new domain to host my projects under.


I think you should be able to do that if you run your own Tangled knot. I'd assume that you can host it under your domain which means your git service will remain even if Tangled itself is down or goes under.


How did people do it before AI? Spellchecks have existed forever. Also, people have the ability to improve, so if you offload all your work elsewhere, you will never improve.


That's an overly optimistic view of things. For a lot of us GitHub is critical infrastructure, which if it goes down loses us and our customers money.


It is interesting, how much money is being lost during this outage?

My significant other was just let go from their job as a scapegoat for an organizational error: 3 layers of failure - IC, manager, director, and the IC was let go. The error caused a 7 figure loss for the company that has 10 figures of revenue per year. The manager and director may not see any consequences, though the director will probably be forced out by end-of-year due to incompetence. The new executive has taken to firing employees much more eagerly than their predecessor, like some sort of Jack Welch acolyte.

Their firing has put a lot of things into perspective for me. Mostly, fuck "at-will" employment and its negative effect on the American social contract.

But also this "angry ranting" online that the original poster was referencing. Not everyone has the privilege to calmly respond to things that directly impact their livelihood.


Any engineered system where an individual can accidentally cause a 7 figure outage is poorly designed. And engineering leadership that decider to terminate an individual due to such failure (as long as there was no malicious actions) is completely clueless.


Captain obvious over here


A story as old as time.


Name names.


Local (to us) firm. Two main reasons I didn't:

  - Have never seen it mentioned on this forum, in any context.
  - We're potentially pursuing legal action.


names or it didn't happen, mate

50% of social media traffic is bots and i want to believe


I'll keep you guys in mind if the lawsuit is successful and we're allowed to publicly speak about it. I believe it's a pretty compelling (and sympathetic) story that deserves to be told. But it's still far too soon for me to be shouting the name from the rooftops.

My SO was fired - this isn't a Google review where we were treated poorly at a restaurant.


that's fair, and ultimately who gives af about online randos, but at the same time skepticism is required


Is responding calmly a privilege to have, or a kindness to yourself?


how is it mutually exclusive


don't do this


I'd be curious what the statistics might actually be for people who are directly affected because their business is suffering vs. people affected because their employer's business is suffering.


Most people's employment prospects are directly correlated with their employers ability to make money.


most people's employers arent firing people over a few hours of github outage.

not to mention that any business which could potentially lose enough money that they would need to let go of developers from a github outage should probably already have some business continuity plans in place.


For Germany it's roughly 1:11 if you go by the self-employment rate (~8% of the workforce).


If its so critical why relying on it, and not having ie some mirror or some other way to handle any sort of outage like this. its not like Microsoft is your friend or good business partner, ever.

With every single of these enterprise 'cloud' offerings you are giving (almost) complete power over your business/project to somebody else who couldn't care less about your success or failure, you are simply irrelevant for them. I see it at work too, every time critical external systems go down whole bank stops still, just because few bucks were saved yearly on some on-prem servers.

Look at it this way, you are learning some important lesson today and finding great area of improvement for resiliency from now on.


> For a lot of us GitHub is critical infrastructure

Please read https://berthub.eu/articles/posts/cyber-security-pre-war-rea...


Boomer opinion but trusting third parties to be critical infrastructure, especially with no SLA in sight, will always end in tears. "The cloud" is very convenient, but its providers will never care about your infrastructure or your customers as much as you will.


Older millennial here and I agree. If you rely on Github you should at least be have your processes such that you can work around it.


That's the reason why my build agent is self-hosted


3P-maintained infrastructure is what makes civilizations work efficiently. We're not all digging our own wells, generating our own electricity, and burning or burying our own garbage.


You are still supposed to be prepared for outages.


Agreed. So is the issue really then that people were inadequately prepared with backup plans and now they're suffering the consequences?

It's not all that different from, say, an AWS region having a service impact. People would rather complain about AWS than prepare and utilize a well-tested recovery plan to shift to a standby region. Oftentimes there's no fallback plan because the business already considered it and decided it was too costly relative to the benefit, but when the incident happens, they still can't help but complain. Humans being humans.


Eeh, you always rely on someone else's infrastructure. Even if you are off cloud entirely and own your own datacenter, you still need peering/transit. You still need power (or at least fuel). You can't avoid depending on a DNS infrastructure.

Github does have an SLA: https://github.com/customer-terms/github-online-services-sla. But like virtually all SLAs, it's really just a token gesture. I have never seen an SLA that pays the losses you suffer due to the outage.


People are already saying things like "We need a plan B in case we urgently need to deploy a fix to production, and GitHub Actions is unavailable again".

But in general, it's not feasible to do everything in house.

And I don't think that GitHub is devoid of SLA: https://github.com/customer-terms/github-online-services-sla

The issue is that they're not achieving two nines uptime in practice.


Boomer here as well, but I'd add that trusting your own org for critical infra usually also ends in tears. Most everything in IT involves failure, including in well designed systems designed by great engineers.

I worked for a few years in an exceedingly well capitalised place which ran everything in their own data centers, money no object, with a truck parked somewhere, ready to go, with a smaller version of our critical infra. We had a serious business-stopping outage once every 18 months or so, every time for fringe reasons one only learns about when trying to run a large data center. Its convenient to blame the cloud and pretend that self-hosting in private sector was so, so great with six nines.


At this point github is barely managing one nine.


True. I'd be very annoyed if I had anything mission critical on github. While we have everything in the cloud, we do actually self-host that part.


+1 (as a millenial) ... especially given that setting up a git server for non-OSS company code isn't too much of a challenge really. also, no need to self-denigrate this reasonable opinion in preemptive obedience.


GitHub is way more than just a git repo host. It manages code reviews, merge (pull) requests, and has an entire CI/CD workflow engine in it. Replicating all that is a challenge that most orgs are not up to.


Agree. But also, it's affecting everyone equally, whether they have a free personal account or are part of an enterprise account with SLA. Understanding the practical value of an SLA is an interesting problem.


You're right.

However, I'm also of the boomer opinion that you should get what you pay for. "Ranting online" about a service (you pay for) being unavailable is a reasonable reaction. It's not like they have a call center you can dial into for support ...


How did this become a boomer opinion? It is proved truth thousand times a day. Not that you shouldn't use third parties - but in this industry you can shrink this exposure to the minimum, and have plan B for anything else.


Boomer who has never had a business? Depending on third party suppliers is business as usual.


Well it's not business at all today!


Arguably, git is decentralized, so contingency plans, like working with local and server repos, should still work.


Truly a boomer opinion. Respect!


A lot of casual tech folks I know don't understand GitHub's role in CI/CD, and think it's "just" revision control storage. It's natural, since many people have no reason to know about systems like ghcr.io and npm registry if they're storing vibe-coded personal projects.


And it's not just the user's fault, GH spent a considerable effort in marketing to position themselves as such, see Github Actions and similar junk.


GitHub Enterprise Cloud has been chugging along with no issues. I hope your critical infrastructure isn't dependent on a free tier / service. And that you have a business continuity process in place.


Are you sure? We have enterprise and I couldn't even access our repo briefly.


We have data residency in the EU and didn't notice anything.


GHE.com has only been around since November 2024. Most GHEC customers I imagine are still stuck on the shared global infrastructure on github.com.


"It's just money. It's made up. Pieces of paper with pictures on it so we don't have to kill each other just to get something to eat". Jeremy Irons in movie Margin Call


You cant eat a quote


but you can download a car


How is github being down losing you guys money?


If you pay developers x money / day and one of their core tools is down for n hours during the day and they spend their money on HN instead that's pretty straight forward to calculate.


I meant more for their customers specifically


we push customer code on mondays. it is monday. code is not being pushed.

we do not bill.

the outages they promised their customers are now different.

this has costs for everyone downstream.


I was in the middle of a hot fix. Our pipeline goes through Github.


That's understandable, but surely there's a way to bypass it? You need to have a breakglass procedures lol


“lol just use a workaround” doesn’t work in an environment with hundreds or thousands of employees coupled with audit, security, and other legal requirements to ship software.


If you're actually bleeding money, you better believe you'll get permissions for a workaround, if you know what you're doing.

It all ties back to the OP, where the issue you've might not be as bad as you think. I have been in situations where we have dropped all procedures to push a hot fix because we were actively bleeding money, and in situations where you know there is an issue, and you let it be.


What an odd hill to die on. Maybe I just lost a $30M client because they wanted an issue fixed now. Maybe my company would get sued for violating their contractual obligations if someone went cowboy to bypass procedures.

Maybe I pay for a service and I want that service to work consistently during core business hours.


"Just recreate GitHub internally."


"just redo the whole CI/CD pipeline bro, it's easy"


Imagine Github being critical infrastructure for you... The ineptitude...


We use github quite a bit as well. Github being down currently is not actively losing us money, i.e. having customer impact.


Fastmail has very nice email sync for iCloud that I use :)


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: