Hacker News Reader

themgton Codex starts encrypting sub-agent prompts

It's sort of insane though, you not only have dozens/hundreds of stochastic agents running on your machine, but you cannot even inspect the instructions those agents are working off of?

I've gone in to look at Claude subagent/workflows and sometimes been like "no this was a mistake to spin up" ... Codex users just get to token yolo the encrypted telephone operator instructions+shell from orchestrator to subagents?

themgton Zig Creator Calls Spade a Spade, Anthropic Blows Smoke

Yet one is technical and the one you actually got a negative opinion about a programming language supposedly isn't?

Yes, on the one hand here's "Structure and Interpretation of Computer Programs" and on the other hand here's my blog "embedding-shape is a stinky slop beginner shit show"

And yet you got a negative opinion of me from the "embedding-shape is a stinky slop hack beginner shit show" post rather than SICP? I feel like I woke up in an alternative universe!

themgton Don't you mean extinct?

> I’ve never been in a situation that says “We’ve not think of that, we need to get back to the drawing board”

And you're somehow able to post on HN from within the time crystal?

themgton Croc: Securely transfer files and folders between two computers

Hard-coded to use his relay server and then asking for donations for bandwidth?

https://github.com/schollz/croc/issues/931

themgton AI 2040: Plan A

OK. Anyhow ... if there's a cognitive task you are personally superior to Fable at, let us know.

themgton AI 2040: Plan A

To meta-unpack a little bit ... it is strange to me that Fable is far more capable of discussing these questions than apparently 99% of humans. Along with being more capable at quite a lot else than most humans.

themgton AI 2040: Plan A

What's strange to me about these comments is they're timeless. They could have been written in 2026 or 2016 or 1966.

Like, afaict, for many on HN going from ELIZA->Fable 5 just didn't cause any update to priors regarding this whole philosophical question. The argument against has remained unchanged. I don't see any point in arguing about it, I just find it very strange.

themgton My thoughts on the Bun Rust rewrite

Zig is getting that Elm, etc vibe. Genius/visionary BFDL who's also personally incapable of leading the project towards healthy long-term viability.

Say what you will about Matz or José Valim, I don't think they'd ever write a "and don't let the door hit you on the way out" screed full of personal attacks ("stinky manager", "writing slop", "a total shit show") against a person who led a very prominent project and financially supported the language.

themgton We Got This Wrong. and We Are Fixing It

HubSpot has been building a different solution for our customers. On August 4, Contact Discovery launches — and for the first time, your team can find, verify, and add net-new contacts without ever leaving HubSpot.

To support that, we’re updating our Customer Terms of Service, Product Specific Terms, Privacy Policy, Sub-Processors Page and Data Processing Agreement, effective July 1, 2026. This post explains what’s changing and why.

The contacts you find through Contact Discovery are reliable because they’ve been checked for deliverability, accuracy, and whether that person is still at that company. You’re not buying a raw list and hoping for the best. Every contact that surfaces has been validated, and decision-makers are ranked first.

That quality is only possible because of a shared dataset. When you opt into enrichment, some of your business contact data helps keep that dataset current. Everyone who participates gets more accurate data back in return. You never pay for a contact already in your CRM, and there’s no separate contract.

These are the new leads. These are the HubSpot leads, data mined from your own HubSpot account. And to you they're gold, and you don't get them. Why? Because to give them to you is just throwing them away. They're for closers. I'd wish you all good luck, but you wouldn't know what to do with it if you got it.

themgton What Emily Bender meant by "stochastic parrots"

What I have been doing in many places—the octopus thought experiment, stochastic parrots, the phrase “synthetic text-extruding machines”—it’s all about trying to make vivid to people who aren’t in the business of building language technology what these systems actually do

> Meanwhile, O, a hyper-intelligent deep-sea octopus who is unable to visit or observe the two islands, discovers a way to tap into the underwater cable and listen in on A and B’s conversations. O knows nothing about English initially, but is very good at detecting statistical patterns. Over time, O learns to predict with great accuracy how B will respond to each of A’s utterances. O also observes that certain words tend to occur in similar contexts, and perhaps learns to generalize across lexical patterns by hypothesizing that they can be used somewhat interchangeably. Nonetheless, Ohas never observed these objects, and thus would not be able to pick out the referent of a word when presented with a set of (physical) alternatives.

This seems kind of obviously wrong at least in the context of coding agents. These models get trained on actual output of the previous version of the model doing its job, often "IRL" on a real computer/project. It's like O is in the conversation for years now and learning from his own interactions between A <-> O <-> B, where A is the human and B is the computer.

The idea O ontologically has never "observed" "these objects" or referents is philosophically strained. Have I observed the moon, or a finger pointing at the moon? Have I observed `sed` more than Fable?

themgton OpenWiki: CLI that writes and maintains agent documentation for your codebase

Yeah I went looking: https://github.com/langchain-ai/openwiki/blob/main/src/agent...

themgton Agentic coding notes from Galapagos Island

Forgetting? I think you mean to say your advice was auto-compacted to keep our context small and deliver better results.

themgton Fable 5 Is Back

afaik there's somewhat painful economics. Not sure back-of-napkin but something like:

   • 150-500B: Sonnet
   • 0.9-2T: Opus
   • 3-5T/10T: Fable / Mythos

So if bigger model is "smarter" but you effectively wind up with a "shared hosting" model where a coherent inherence node(s) that cost $2m or something can run max 10x customer workloads simultaneously ... not sure what that can be priced at.

If it turns out a $10m/10x shared node can host even smarter models, then what?

themgton You can't unit test for taste

This is exactly it - the ultimate skill now is to be Rick Rubin with an LLM. Not a comfortable transition as a coder.

themgton Ask HN: Where is the programming profession going?

"Well, in our country," said Alice, still panting a little, "you'd generally get to somewhere else—if you ran very fast for a long time, as we've been doing."

"A slow sort of country!" said the Queen. "Now, here, you see, it takes all the running you can do, to keep in the same place. If you want to get somewhere else, you must run at least twice as fast as that!"

themgton Ask HN: Where is the programming profession going?

I will just say, if you are any good at programming and have experience using agents, you're in the top 0.1% of the world in adoption of a critical new technology.

It may seem hopeless as a programmer, but imo you'd be much better off reframing your situation re: the above sentence.

themgton GLM-5.2 is a step change for open agents

I just tested GLM 5.2 out via Z.ai in pi for a little one-off project that was already scoped. It actually did a relatively decent job starting out, and figured important things out from context.

But the reasoning traces became increasingly hilarious, with it getting confused and going in loops, doubting itself. I began to feel almost sad, it was like listening to the internal monologue of someone with anxiety disorder.

It made pretty good progress but wound up going in a lot of goofy loops and doing things a bit "off" from standards I'd hoped it would infer, and finally started going a bit nuts, "This is very confusing.", "OH WAIT", seemingly hallucinating a whole side-quest that didn't make sense and looking at making internal system changes to try to achieve its (now very confused) goal when I pulled the plug.

Without seeing the reasoning traces from Claude/GPT it's hard to really know, but it definitely didn't feel like the same quality of reasoning, even if dogged persistence does wind up actually working eventually.

themgton Jobs and Software Is Fucked

It's not like the job market was that much better before AI infested every single corner of the market, but it supercharged all of the worst aspects of everything. I've seen people supposedly smarter than I advocate for just giving in, conceding to AI coding as it's the future. But doing so means tossing out my friends who make art or the people who work their asses off to properly test and review code or the writers pouring all of their energy into even mundane dialogue. It means throwing out my dignity as a software engineer, as someone that truly gives a shit about security and code.

Don't let yourself get attached to any tech stack you are not willing to walk out on in 30 seconds flat if you feel the heat around the corner. That's the discipline.

themgton GPT-5.5 hallucinates 3x more than MIT-licensed GLM-5.2

instead of people’s vibe checks and pelican SVGs.

Right, what happened is everyone went to Fable and asked it to make the very best bicycle pelican SVG, no mistakes. And Fable's bicycle pelican SVGs were such timeless masterpieces, we all instantly got AI psychosis. Happily, you were immune to this.

themgton From PGP to Mythos: a brief history of export controls that didn't stop anyone

This was never the question. The question was, will the export controls slow them down in the short/medium term to the point where it will give US companies an advantage?

I mean, I remember listening to the Biden people back in 2022 talking how they were going to cripple China's semis and therefore AI industry and keep them 5+ years behind the curve as Team America accelerates ahead. That was the pitch.

You've now got Huawei Ascend 950, GLM-5.2 at Opus 4.8 levels, China dominating OSS models, and Z.ai saying they'll have a Fable-level model by EOY. I would say the export controls have utterly, utterly failed.

https://www.csis.org/analysis/understanding-biden-administra...

themgton GPT-5.5 hallucinates 3x more than MIT-licensed GLM-5.2

That’s not what your quotes said. They said bigger models = plateau in intelligence, nothing about more data or increased hallucinations ... I’m pretty sure #1 is well known

Well known in a multiverse branch where Fable was a dud?

themgton RFC 10008: The new HTTP Query Method

Great, got it. I'll update my running "how computing works" chart with this new information:

  | implementation = reality | magic  |
  |-----------------------------------|
  | 999,999,999,971 (+1)     | 0      |

themgton Local Qwen isn't a worse Opus, it's a different tool

That said, I can't wait for LLMs to stop being AI and start being just another tool.

From a horse's perspective, the internal combustion engine is just another tool for making scary noises and powering horse trailers to take me on fun horse adventures. So ... perhaps.

themgton RFC 10008: The new HTTP Query Method

Putting something in a spec does not automatically make it true.

Terrible news for computing.

themgton Launch HN: Adam (YC W25) – Open-Source AI CAD

Just to self-promote, we've got a very early stage project with a lot of similar ideas at https://quidities.com/ - feel free to signup / reach out if you have real world use cases anywhere in this space - we've got a lot in the works already.

themgton GPT‑NL: a sovereign language model for the Netherlands

This "vaguely nationalistic world view around tech" is a direct consequence of the US government ... The EU was build on the principles of collaboration ... But this only works if it is reciprocal.

This sounds great but doesn't really make any sense. What skeptics are saying is there should be a pan-EU effort to build frontier models, rather than one-off toy models built by each country as a box-ticking exercise.

"We built the EU as a powerful supranational organization, so logically the EU's response to a great power challenge in the realm that may well define the 21st century is gonna be to shatter its efforts into 27 useless pieces, because Trump bad" is just absolutely ridiculous and will not lead to anything good. Be the change you want to see, etc.

themgton Noise infusion banned from statistical products published by Census Bureau

My favorite gerrymandering story was when I learned an ~800 student local college had been split down the middle, so students were in different congressional districts depending on their dorm building.

themgton Amazon CEO's talks with U.S. officials triggered crackdown on Anthropic models

I submitted separately, but this Axios report has some details that call a lot of the speculation in this thread into question, i.e. that this wasn't much of a "jailbreak" at all and that it's not Anthropic-specific - the White House intends to generally regulate Mythos-class models (whatever exactly that means):

Between the lines: The government's response "seems way out of line with what's actually in the research report," Luta Security CEO Katie Moussouris, who Anthropic shared the Amazon report with, told Axios.

Moussouris said the researchers were able to find security vulnerabilities by asking questions normal defenders would ask AI, which is exactly what the model was intended to do.

An administration official told Axios they do not view other models as national security threats because they do not surpass the bar that Mythos set.

Anything at Mythos level or above would need to go through the administration to ensure the government's national security apparatus is hardened enough, the official added.

https://www.axios.com/2026/06/13/anthropic-amazon-white-hous...

themgton Statement on US government directive to suspend access to Fable 5 and Mythos 5

I think we should see this as simply silly behavior by a government. ... We saw some of this behavior in the last administration too.

So it's silly behavior, as typified by the last decade of American governance? Is there "serious" American leadership we should be expecting to see soon, e.g. 2029 AOC elected on a platform of unlimited 10GW datacenters and universal basic Mythos 8 models?

It may seem subjectively silly to you, but e.g. getting executed for refusing to point at a deer and call it a horse is pretty silly stuff as well, at least for those not living in the Qin Dynasty.

US voters deserve better.

Deserve's got nothing to do with it.

themgton I Am Not a Reverse Centaur

Some of this is the funny situation where the faithful will state: "This writes better code than I do!" and miss the irony of: "yes, yes it does"

"Blessed are the humble ..."