For context, two days ago some users [1] discovered this sentence reiterated throughout the codex 5.5 system prompt [2]:
> Never talk about goblins, gremlins, raccoons, trolls, ogres, pigeons, or other animals or creatures unless it is absolutely and unambiguously relevant to the user's query.
They saw how Google providing absolutely terrible customer service for a very long time has done nothing to hurt their bottom line and decided to copy.
I used to work at GitHub. I think you should find a new job.
Before Microsoft came along, the entire company was aligned from the bottom to the top around the goal of delivering a single great product. As soon as they bought us, that changed; there were now lots of ways for an individual to succeed at GitHub-the-division-of-Microsoft that had nothing to do with GitHub-the-product. Now GitHub doesn't even have its own top, the org chart just smears into the Microsoft one at some hazy point. Perverse incentives abound.
An organization like Microsoft can never recreate the magic that was GitHub. There's just too many competing interests and agendas that have absolutely noting to do with making GitHub better. In the time before I left, I actually encountered many people who didn't care if they were making it worse, if it advanced their other goals.
After going public and getting publicity. You shouldn't have to do that just to get a company to fix their own mistake. They stole $200, where do they get off saying they won't give it back?
We are slowly inching closer to the point where AI and AI products will be billed for what they cost. We are currently living in the heavily discounted world where everything subsidized to the point where a lot of it is free. It seems like they can't or won't keep that up anymore. My prediction is that whenever one of the big companies raise their prices or move features to higher tiers others will follow soon. They all feel the pressure and non of them want to give away more money than they need to.
I wonder if managers will be as excited about AI when the prices go up.
This post comes uncomfortably close to plagiarizing https://thebuild.com/blog/2026/04/23/preempt_none-is-dead-yo..., which it cites as a source; almost all the technical explanation is in there and some of the wording is extremely similar. Compare, e.g., "What Linux 7.0 actually changed" in Pettus's post to "What Is Preemption?" in this one. I think this link should have been to Pettus's post instead.
What did Republican appointees on the Supreme Court have to say about student loan forgiveness?
Who not only fights tooth and nail against healthcare improvements but actively took money from existing programs to fund a national police and detainment apparatus?
Which party produced the most meaningful (albeit not far enough) healthcare reform of the 21st century in the US?
Those examples in particular are quite rough to try and both sides.
I feel like MS went out of its way to make a point that GitHub and NPM would be independent orgs that no longer had to worry about making keep-the-lights-on money. It was positioned as a benevolent acquisition for the good of the development community.
As so often happens, that didn't last long.
Nest was originally independent. Didn't take long for it to merge with the Google Home brand.
Anthropic doesn't even use their own harnesses for their support chatbots (they're using fin.ai) - that's how little support matters to them. Seems like either you get attention on HN, know someone working there, or are at a large enough company to have an enterprise contact - otherwise, no reply.
FWIW, the top comments at the time of my comment (one hour after yours, two hours after the article was posted) are all complimentary. You commented one hour after the article was posted; it's worth waiting a bit for the comment voting to shake out.
THe government shouldn't be raising anyone's children, that's what parents are for. If you're a bad parent, your kids will get access to bad things and could become an adult failure.
The future of your family and your legacy is up to you, not the government. We don't need age verification to restrict the social darwinism of raising children.
I think the big secret is that AI is just software. In the same way that a financial firm doesn't all of sudden make a bunch of money because Microsoft shipped an update to Excel, AI is inert without intention. If there's any major successes in AI output it's because a person got it to do that. Claude Code is great, but it will also wipe out a database even though it's instructed not to (I can confirm from experience). The idea that there's some secret innovation that will come out any minute doesn't change the fact that it's software that requires human interaction to work.
Would love if OpenAI did more of these types of posts. Off the top of my head, I'd like to understand:
- The sepia tint on images from gpt-image-1
- The obsession with the word "seam" as it pertains to coding
Other LLM phraseology that I cannot unsee is Claude's "___ is the real unlock" (try google it or search twitter!). There's no way that this phrase is overrepresented in the training data, I don't remember people saying that frequently.
It seems there was some kind of confusion during the disclosure process, because the vendors aren't treating this vulnerability as serious and it remains unpatched in many distros.
I agree with the article, FastCGI is better than HTTP for these things.
Though I'd like to make another protocol known: Web Application Socket (WAS). I designed it 16 years ago at my dayjob because I thought FastCGI still wasn't good enough.
Instead of packing bulk data inside frames on the main socket, WAS has a control socket plus two pipes (raw request+response body). Both the WAS application and the web server can use splice() to operate on a pipe, for example. No framing needed. Also, requests are cancellable and the three file descriptors can always be recovered.
Over the years, we used WAS for many of our internal applications, and for our web hosting environment, I even wrote a PHP SAPI for WAS. Quite a large number of web sites operate with WAS internally.
How will this end up going any better than Mastodon has?
Near inevitabilities:
- All the small instances defederating from the largest due to politics/spam/annoying noobs/whatever, effectively killing the easiest path to entry into the community
- Pointless debates about whether it’s OK to federate with instances that host pirated content, disagreeable politics, furry VNs, etc., which everyone has to take a side (the correct side) on
- Relatively little actual work/productive discussion going on, since many users are there mostly for the politics / fediverse posturing than for actual work
Absolutely, in heaps. The very moment you start Warp, it sends 5 HTTP requests, (1 version check, 1 LLM model list, 3 telemetry events). The former two go to app.warp.dev, the latter 3 go to warpianwzlfqdq.dataplane.rudderstack.com and include a persistent UUID, your operating system and version, Warp version, and the tracked event name and its properties.
All of this happens before you even see the window. After you're done clicking 'No' to all the SaaS nags, you can turn off telemetry in the settings, but for some reason it gets turned back on when you restart the terminal.
While the terminal is running, it calls home whenever you trip one of the events listed in this 7000-line long file: https://github.com/warpdotdev/warp/blob/d0f045c01bacbd845a63.... Besides all the hosts listed earlier, it also makes requests to o540343.ingest.sentry.io.
> The US Copyright Office confirmed this in January 2025, and the Supreme Court declined to disturb it in March 2026 when it turned away the Thaler appeal. Works predominantly generated by AI without meaningful human authorship are not eligible for copyright protection, and that rule is now settled at the highest judicial level available.
Misstates the law. Denial of certiorari can happen for many reasons unrelated to the merits and does not settle the issue nationwide.
It’s just an impossible problem. Photons don’t provide sufficient information to determine calories (at least not in any way they could practically be captured). Inside that sandwich could be drenched with olive oil or it could be hollow cheese with lettuce. It’s impossible to tell.
It's an odd thing here, because I don't really understand why this is LLM-specific at all. If someone came up to me and asked "who's the 6 Nimmt world champion?" I'd google it and probably find the same result, and have no reason not to believe it. I mean, for all I know the game is being made up too, though it has more sources at least.
Aren't you forgetting the part that says "solely: (a) to perform its obligations set forth in the Terms, including its Support obligations as applicable; (b) to derive and generate Telemetry (see Section 4.4); and (c) as necessary to comply with applicable Laws. Except as required by applicable Laws, Zed will not provide Customer Data to any person or entity other than Customer’s designees (including pursuant to Section 7) or service providers."
Seems like legalese to be able to take that data for support reasons, telemetry, and local laws that require that data be sent to them. I think ignoring this portion is a little uncharitable to them.
For context, this is coming in as TriMet is laying off staff, reducing service frequency, eliminating bus lines, and cutting parts of light rail routes due to a $300M budget shortfall. The cuts were exacerbated by state Republicans getting a proposed payroll tax repeal onto the ballot next month; TriMet relies heavily on payroll taxes that are deeply unpopular among the self-employed and small business owners, so the budget is going to get worse before it gets better.
So at the same time that public transit is retreating and rideshare company labor overhead is threatening to increase, Waymo shows up with a convenient solution to both problems.
Is this microsoft stating that they aren't able to get acceptable reliability from Azure? (I mean, I think a lot of us have heard that, but it's interesting to hear it from microsoft themselves).
Have you looked at the results for any commercial query, something like [sofa beds] or [hard drives]? It is basically 100% ads. Anything where the user is intending to spend money, they show only ads, and have all the top producers in the world bid against each other for who gets featured, and Google captures essentially all surplus value in the transaction.
My wife is an investor, and one of her portfolio areas is pharmaceuticals. A couple of portfolio companies have reported that it's becoming basically impossible to make any money off of a new product, because you need to advertise it to reach the customer, and Google will skim all the excess producer surplus off as you compete with other startups serving the same market.
It's basically the perfect business model. They own the path to the consumer, which means they own the economy.
I'd also recently hired someone out of Google Search, and they said that the only queries that "legacy" (non-AI-mode) search cares about are commercial-seeking queries, and the only metric they optimize for is ad conversions on those. It literally is thousands of people whose only job is to get you to click more ads.
> Never talk about goblins, gremlins, raccoons, trolls, ogres, pigeons, or other animals or creatures unless it is absolutely and unambiguously relevant to the user's query.
[1] https://x.com/arb8020/status/2048958391637401718
[2] https://github.com/openai/codex/blob/main/codex-rs/models-ma...