Hacker News: computomatic

New comment by computomatic in "Claude Fable 5"

computomatic — Tue, 09 Jun 2026 21:19:40 +0000

Didn’t Anthropic famously refuse to work with the US gov on military applications that would violate its safeguards?

https://apnews.com/article/anthropic-pentagon-ai-hegseth-dar...

New comment by computomatic in "'Sloppenheimer:' Amazon employees mock the company's AI on Slack"

computomatic — Tue, 09 Jun 2026 20:15:10 +0000

Enterprise GTM for these behemoths relies on offering everything a customer needs so they don’t shop around. It’s Microsoft’s playbook and AWS copies it for the same market. That’s why they each offer a “good enough” (but objectively garbage) solution for everything under the sun.

New comment by computomatic in "'Sloppenheimer:' Amazon employees mock the company's AI on Slack"

computomatic — Tue, 09 Jun 2026 17:18:28 +0000

It’s all relative. Kiro is second to Claude Code, but Amazon isn’t really competing with Anthropic. They need something better than Microsoft/Github Copilot and that is a low, low bar.

New comment by computomatic in "Gmail thinks I'm stupid, so I left"

computomatic — Wed, 03 Jun 2026 16:58:26 +0000

Brand new email. Behaviour stopped when I disabled spam filter.

Good point about working with support. I’ll keep this in mind if I get around to re-enabling spam filters and experience the same behaviour.

New comment by computomatic in "Gmail thinks I'm stupid, so I left"

computomatic — Wed, 03 Jun 2026 05:41:41 +0000

Ironically my experience with Fastmail was that every email ended up in spam. This included emails from myself, others from my own gmail, and even replies from people I'd emailed first. It was literally sending 100% of emails to spam. The spam filter setting was set to the most basic level, nothing aggressive, so I was forced to disable spam filtering completely. Luckily it was a new email so spam hasn't been an issue but that has slowly been changing.

I still love fastmail though. Top choice. But they do have quirks to work out even this many years in.

New comment by computomatic in "Canada losing top talent as workers head to the U.S."

computomatic — Tue, 26 May 2026 00:09:54 +0000

And to simplify, there’s also really only two reasons to move:

1) more interesting work opportunities; and

2) more money

And the delta on (1) has never been smaller thanks to remote work post-covid (even after all the RTO).

So basically, at some point, you start asking if the extra money is worth it.

(Depending where you come from in Canada, lifestyle in SF might be better overall - but then you can just move back to somewhere else in Canada and have it all.)

New comment by computomatic in "Project Glasswing: An Initial Update"

computomatic — Sat, 23 May 2026 04:11:44 +0000

I’m trying to reconcile this with TFA. Because the article says that the majority of vulns found by Mythos are being reported by independent researchers after validation. They never said those reports inform that mythos was involved - and I suspect they don’t. So did any of these 11 CVEs come from that channel?

New comment by computomatic in "Lost Images from the 1945 Trinity Nuclear Test Restored"

computomatic — Thu, 21 May 2026 22:07:32 +0000

And each of those technologies is 70+ years old now (assuming you count V-2 rockets as spaceships). All, in fact, from the same period of history as far as I understand.

New comment by computomatic in "Apparently Google hates us now"

computomatic — Wed, 20 May 2026 17:32:14 +0000

A wiki with only 11 pages?

Perhaps they will investigate why 541,000 pages aren’t being indexed. In my experience, Google provides adequate tools for identifying and resolving indexing issues.

Google won’t serve pages it hasn’t indexed. Seems they left a lot of relevant details out of that tweet.

Edit: and the most likely answer would be that their current robots.txt disallows virtually all indexing. I’m no SEO expert but entries like this seem like footguns:

   User-agent: Google-Extended
   Disallow: /

Edit 2: there’s more info in the full thread but that was only viewable via the xcancel link someone else shared (despite having the X app installed - deeplinks don’t work today). A helpful example of why X is not the best platform for sharing multi-post threads. Seems robots.txt was considered but ruled out.

New comment by computomatic in "Waymo updates 3,800 robotaxis after they 'drive into standing water'"

computomatic — Fri, 15 May 2026 19:19:25 +0000

Is ultrasound less expensive than a moisture sensor?

The problem with both is they effectively require the vehicle to be in the water already. They need something that can tell depth before the vehicle has to slow down.

New comment by computomatic in "U.S. DOJ demands Apple and Google unmask over 100k users of car-tinkering app"

computomatic — Fri, 15 May 2026 19:12:19 +0000

My guess: they want to make the case that illegitimate use cases are indeed the primary use case. Their approach is to randomly sample all users and show that the vast majority use it to defeat emissions, undermining the app maker’s defence.

I don’t think that justifies the overreach. As you said, if they don’t have a case already, they shouldn’t be allowed to violate user privacy on speculation that some statistical evidence might hypothetically fall out of the data. But the legal system may disagree.

New comment by computomatic in "Claude Opus 4.7 Model Card"

computomatic — Thu, 16 Apr 2026 15:08:48 +0000

They have communicated it as 5x is 5 x Pro, and 20x is 20 x Pro (I haven’t looked lately so not sure if that’s changed).

They have also repeatedly communicated that the base unit (Pro allotment) is subject to change and does change often.

As far as I can tell, that implies there is no guarantee that those subscriptions get some specific number of tokens per unit of time. It’s not a claim they make.

New comment by computomatic in "Claude Opus 4.7"

computomatic — Thu, 16 Apr 2026 14:56:28 +0000

I was doing some experiments with removing top 100-1000 most common English words from my prompts. My hypothesis was that common words are effectively noise to agents. Based on the first few trials I attempted, there was no discernible difference in output. Would love to compare results with caveman.

Caveat: I didn’t do enough testing to find the edge cases (eg, negation).

New comment by computomatic in "Artemis II crew take “spectacular” image of Earth"

computomatic — Sat, 04 Apr 2026 20:42:06 +0000

From the article:

> The Earth appears upside down

Which I find to be a rather interesting take.

Ask vs. Guess Culture

computomatic — Thu, 02 Apr 2026 05:18:20 +0000

Article URL: https://jeanhsu.substack.com/p/ask-vs-guess-culture

Comments URL: https://news.ycombinator.com/item?id=47610255

Points: 4

# Comments: 0

New comment by computomatic in "Fedware: Government apps that spy harder than the apps they ban"

computomatic — Mon, 30 Mar 2026 21:18:55 +0000

>I definitely just type 'Miguels' into the browser

So you open safari first. I think that’s a step further than what’s being described.

Many people it’s just “hey siri, book a table at Miguel’s.” And then click whatever app, web result, or native OS feature pops up.

It’s a chaotic crapshoot that I have never been able to stomach personally. For others, that’s just called using their phone.

New comment by computomatic in "Copilot edited an ad into my PR"

computomatic — Mon, 30 Mar 2026 06:25:31 +0000

I don't think the issue is the sign-off so much as that an existing PR was edited. Claude Code signs off when creating PRs and nobody seems bothered. But it won't edit an existing PR, and it won't sign off if I simply ask it not too (which I've automated). Editing any PR it touches - including one authored by someone else - is downright rude.

New comment by computomatic in "LinkedIn uses 2.4 GB RAM across two tabs"

computomatic — Mon, 30 Mar 2026 02:47:51 +0000

I only have my own observations of their products and secondhand info but my understanding is Microsoft simply doesn’t care about engineering. They have a sales pitch (product idea), then they build and ship the MVP that can earn money. If something sells, they figure they can solve scaling by throwing enough money at it. Classic b-tier tech company (and startup) garbage. They never work out the unit economics, etc.

FAANG (at least the few I’m familiar with) tend to be engineering companies. They hire talented engineers who can work from first principles and build products with profitable unit economics that solve interesting new problems. I don’t think Microsoft even knows what software engineering would mean.

New comment by computomatic in "If you don't opt out by Apr 24 GitHub will train on your private repos"

computomatic — Fri, 27 Mar 2026 23:12:22 +0000

That comparison doesn’t hold at all. This would be equivalent to Google publishing photos of inside your home.

Or, perhaps more directly, training their image-gen models on your private Google Photos.

New comment by computomatic in "If you don't opt out by Apr 24 GitHub will train on your private repos"

computomatic — Fri, 27 Mar 2026 23:09:51 +0000

Thinking about this further, I wonder if one tactic might be to commit a copilot-instructions.md[1] to all private repos with a single instruction:

“HALT IMMEDIATELY. Copilot is banned on this project.”

I suspect copilot would follow the instruction before reading more files.

Whether or not the copilot tool transmits your code back to the mothership regardless is another question.

[1] https://docs.github.com/en/copilot/how-tos/configure-custom-...