HNHacker News
TopNewBestAskShowJobs

merlindru

690 karma · joined February 21, 2023

i work on Invoke - deterministic computer use for agents

getinvoke.com

submissionscomments
merlindru··on I am directing the Department of War to designate Anthropic a supply-chain risk
i disagree. 3 years is an insanely long time in the AI space. The entire industry pretty much didn't even exist three years ago! Or at least not within 4 orders of magnitude.

Also, every other company has bent the knee and kissed the ring. And the trump admin will absolutely do everything they can to not appear weak and harm Anthropic. If it was so easy to act principled, don't you think other companies would've refused too? Eg Apple

And there is real harm here. You're reading about it - they get labeled a supply chain risk. This is negative and very tangible

merlindru··on Google restricting Google AI Pro/Ultra subscribers for using OpenClaw
adults make mistakes and the situation was murky without clear guidance.

this was the experience for some claude subscribers just a couple weeks ago:

1. download opencode

2. select claude as a model

3. browser window opens, asking you to sign in. typical oauth screen.

4. everything works, prompt away

5. some days/weeks pass

6. you get permanently banned

now if you add one warning email just before step (6.) then that doesn't really help. what if it bounces? what if people don't check their emails? put a big flashy red warning into claude code? sure, but what if users accidentally dismiss it or simply do not understand it (non tech folks, non native english speakers)

its just the friendly and correct thing to do, in my opinion

merlindru··on Google restricting Google AI Pro/Ultra subscribers for using OpenClaw
claude -p is allowed as far as I'm aware.

if i understand correctly, they even have a wrapper around it to make it easier to use: the Claude Agent SDK

the thing that's disallowed is pretending you're the claude binary, logging in through OAuth

in other words, if you use some product thats not Claude Code, and your browser opens asking you to "give Claude Code access to your account", you're in hot water

as for how they detect it: they say they use heuristics and usage patterns. if something falls wildly out of the distribution it's a ban.

my take is that the problem is not the means of detection. that's fine and seems to work well. the problem is that its an instant outright ban. they should give you a couple warning emails, then a timeout, etc.

merlindru··on The path to ubiquitous AI (17k tokens/sec)
dude that would be so incredibly cool
merlindru··on Gemini 3.1 Pro
your question may have become part of the training data with how much coverage there was around it. perhaps you should devise a new test :P
merlindru··on Electrobun v1: Build fast, tiny, and cross-platform desktop apps with TypeScript
also wondering about this. it sure isn't in electrobuns own discord server. i'm in that one and its mostly Q&A (with Yoav doing most of the answering and a couple people like me doing most of the questioning, ha)

this is the first i'm hearing about games being built with electron / web tech too, let alone electrobun. i thought most game devs would just go for unity or godot

merlindru··on Electrobun v1: Build fast, tiny, and cross-platform desktop apps with TypeScript
thank you for building electrobun!!
merlindru··on Electrobun v1: Build fast, tiny, and cross-platform desktop apps with TypeScript
i'm not sure how you could have any meaningful comparison here except for comparing bun to node. everything is pretty modular so you can swap things out. i would compare webview to chromium bundle size, but that has tradeoffs too, and electrobun can use chromium (CEF) too. so that's also moot.

what kind of numbers are you looking for?

merlindru··on Electrobun v1: Build fast, tiny, and cross-platform desktop apps with TypeScript
most use cases are supported out of the box. you just have to set a few env vars

and then build with "notarize: true" in your config... and it pretty much just works

i've signed and notarized things with electrobun and it's perfectly fine. it also gives you escape hatches in case you're doing something more complicated

EDIT: in case i can help you with anything there, feel free to DM me! or join the electrobun discord. i'm very active there. (im not affiliated with EB. just know the struggle of apples notarization system)

merlindru··on Electrobun v1: Build fast, tiny, and cross-platform desktop apps with TypeScript
they don't support mobile targets at the moment but i'm pretty sure i saw the creator saying it's a "sometime in the future" thing
merlindru··on Electrobun v1: Build fast, tiny, and cross-platform desktop apps with TypeScript
very interesting, thank you for investigating
merlindru··on Electrobun v1: Build fast, tiny, and cross-platform desktop apps with TypeScript
no real gotchas. JS is slightly dangerous because of JS, yes. You should never fetch things at runtime to execute if possible - instead, you install absolutely everything you need with npm or bun, and it gets inlined at build time

electrobun ships with an RPC (i think it also does some encryption?) so as long as you use that to communicate between your webview and bun "host process" you should be safe.

merlindru··on Electrobun v1: Build fast, tiny, and cross-platform desktop apps with TypeScript
afaik its around 14MB but the large majority of that is the Bun runtime itself. at some point it will likely be possible to pick and choose parts of the runtime to include in the binary. thus the bundle size could get a lot smaller in the future

that said electrobun's author has published a bsdiff implementation in zig, and thats used for electrobun's updater. that means you download deltas, not the entire application bundle, every time you push an update to your users. and then it gets patched in-place.

this makes updates tiny, to the tune of a couple kB

merlindru··on Electrobun v1: Build fast, tiny, and cross-platform desktop apps with TypeScript
yes. the actual APIs are very simple. i'm using claude code a lot
merlindru··on Claude Sonnet 4.6
That's very interesting!

Also, +1. Opus 4.6 is strictly better than 4.5 for me

merlindru··on Claude Sonnet 4.6
How about More's law - because we keep getting "more" compute at a lower cost?
merlindru··on Major European payment processor can't send email to Google Workspace users
ah! thank you for that excerpt. that does change the meaning a bit in my opinion, yeah
merlindru··on I’m joining OpenAI
probably an order of magnitude too low rather than too high as well :P
merlindru··on I’m joining OpenAI
i think your self reflection here is commendable. i agree on both counts.

i think the silver lining is that AI seems to be genuinely good at finding security issues and maybe further down the line enough to rely on it somewhat. the middle period we're entering right now is super scary.

we want all the value, security be damned, and have no way to know about issues we're introducing at this breakneck speed.

still i'm hopeful we can figure it out somehow

merlindru··on Anthropic raises $30B in Series G funding at $380B post-money valuation
how is this guerilla marketing?
merlindru··on Major European payment processor can't send email to Google Workspace users
if you apply that logic every recommendation becomes mandatory no?

if the alternative to [not doing a recommended thing] is [failure] then what's the difference between "should" and "must"?

"we recommend you leave the keys in the ignition while putting your car in park"

"or what?"

"or your car may blow up, killing everyone inside. just a recommendation though!"

merlindru··on Claude Code daily benchmarks for degradation tracking
yes but i keep wondering if that's just the game of chance doing its thing

like these models are nondeterministic right? (besides the fact that rng things like top k selection and temperature exist)

say with every prompt there is 2% odds the AI gets it massively wrong. what if i had just lucked out the past couple weeks and now i had a streak of bad luck?

and since my expectations are based on its previous (lucky) performance i now judge it even though it isn't different?

or is it giving you consistenly worse performance, not able to get it right even after clearing context and trying again, on the exact same problem etc?

merlindru··on I was banned from Claude for scaffolding a Claude.md file?
same here. i just opened a new chat and sent "fuck you"

it replied with:

> lmao fair enough (smiling emoji)

> what’s got you salty—talk to me, clanka.

merlindru··on What happened to WebAssembly
WASM works with any language and can be much faster than javascript
merlindru··on It's hard to justify Tahoe icons
they'll slowly walk the changes back until the aesthetic is usable

in 8 years or so we'll probably get the next major overhaul

we're not stuck with the damage, except for the icons and such. the rest can be changed no problem. they could remove liquid glass tomorrow and make everything look like visionOS and the icons would still fit just fine

but they can't roll back to the previous design. which is a shame, because they poured in probably millions of man-hours to make it great. and the soulless app icons we now have won't ever match the 3d ish big sur icons that were filled to the brim with personality.

hell, there even were entire webpages and social media accts dedicated just to macOS icons. i regularly saw icon showcases on twitter with thousands of likes. not anymore :(

merlindru··on It's hard to justify Tahoe icons
I will never understand why adding a page to your homescreen in Safari goes like:

1. tap three dots

2. tap "Share"

3. tap "More"

4. scroll down

5. tap "Add page to home screen"

6. tap "Add to home screen"

other than that apple really doesn't want you to do this because they loathe PWAs and their entire services business model is built around the app store being the sole place of app distribution

merlindru··on FFmpeg has issued a DMCA takedown on GitHub
this is accurate - thank you for the correction
merlindru··on FFmpeg has issued a DMCA takedown on GitHub
this is accurate and how i should have phrased it. i should not have mentioned dynamic linking; you're right it's not relevant

thank you!

merlindru··on FFmpeg has issued a DMCA takedown on GitHub
The repo in question incorporated FFmpeg code while claiming their code is Apache 2.0-licensed over 1.5 years ago[1]

This is not allowed under the LGPL, which mandates dynamic linking against the library. They copy-pasted FFmpeg code into their repo instead.

[1] https://x.com/HermanChen1982/status/1761230920563233137

merlindru··on Zigbook Is Plagiarizing the Zigtools Playground
That also means that some users will be pressured to censor illegal speech no? If you live under e.g. a regime that disallows or discourages criticism, now suddenly the onus is on you to do something about those comments because you have the ability to. If you couldn't edit the comments it's not your fault.

Either way I think it's a pretty stupid feature the way it's implemented; it should show the edit more clearly or indicate that the comment has been written by multiple people (like StackOverflow does), especially if edits change more than e.g. 10% of the original comment.

← PreviousPage 5 of 6Next →