frontpage.
newsnewestaskshowjobs

Open Source @Github

fp.

AI agent can move money in Stripe

https://stripe.dev/blog/your-ai-agent-can-move-money-in-stripe
1•ltononro•2m ago•2 comments

Grimes Joins Cast of Fully AI-Generated Movie 'Gods Don't Give Gifts'

https://gizmodo.com/grimes-shows-off-her-acting-chops-in-new-ai-movie-or-does-she-2000835869
1•bookofjoe•2m ago•1 comments

Chainalysis says breach exposed password hashes and live tokens

https://cointelegraph.com/news/chainalysis-cyberattack-password-hashes-live-tokens-exposed
1•6thbit•2m ago•0 comments

Show HN: A tool for looping and transcribing YouTube Videos

https://looptranscribe.com
1•javier123454321•2m ago•0 comments

Show HN: Infinity – Git-style branching for AI chats

https://infinitychat.app/
1•nkotov•2m ago•0 comments

Show HN: noise'r'us, like placekittens.com but for audio

https://noiserus.greg.technology/
1•gregsadetsky•3m ago•0 comments

The Harness: Deterministic Systems

https://klaviyo.tech/inside-the-harness-deterministic-systems-476413ae913b
1•tremdog•6m ago•0 comments

Addressing Corrupt Uses of Artificial Intelligence (2025) [pdf]

https://files.transparencycdn.org/images/2025_WorkingPaper_Addressing-Corrupt-Uses-of-Artificial-...
1•yubblegum•7m ago•0 comments

Scriptable GeoDNS: Leaving AWS for Bunny

https://vincent.bernat.ch/en/blog/2026-bunny-geodns
1•sshah03•7m ago•0 comments

Promethevs: Full DeepSeek-v4.1-Flash-1017, locally, on 12GB of VRAM

https://github.com/Magnvs7/Promethevs
1•ar-nelson•7m ago•0 comments

WebAssembly Specification Release 3.0 (2026-11-04)

https://webassembly.github.io/spec/core/
2•Tepix•8m ago•0 comments

The United States of Inference

https://subconscious-systems.github.io/united-states-of-inference/
1•Jackobrien•9m ago•1 comments

Six pallets of Nvidia GPUs worth $11.4M stolen in California in Aug 2026

https://www.tomshardware.com/pc-components/gpus/six-pallets-of-nvidia-gpus-worth-usd11-4-million-...
1•paaloeye•10m ago•0 comments

Devin Review 2.0

https://devin.ai/blog/devin-review-2
1•gmays•11m ago•0 comments

I tried to connect every GPU on the planet together (kinda)

1•lukefrisker•11m ago•0 comments

Democrats have the votes to make the Electoral College obsolete

https://therealnews.com/democrats-finally-have-the-votes-to-make-the-electoral-college-obsolete
3•vrganj•12m ago•0 comments

The Formula for Lovable Cities

https://worksinprogress.co/issue/the-forgotten-formula-for-lovable-cities/
1•rafaelc•12m ago•0 comments

AI models built outside the US and China are nearly a year behind

https://village.dev/blog/outside-the-us-and-china-the-world-is-building-last-years-ai/
2•kevin_hu•12m ago•0 comments

LeanPRM

https://twitter.com/rodney_lafuente/status/2118736694937629167
1•andes314•12m ago•0 comments

The Comprehensive Guide to the Bank of Nvidia

https://hntrbrk.com/investigations/nvidia
1•mxschumacher•12m ago•0 comments

China's Connected Industrial Ecosystem

https://michaelhillaert1.substack.com/p/chinas-connected-industrial-ecosystem
1•michaelhillaert•13m ago•0 comments

$11.4M of Nvidia Chips Vanished in Oakland

https://tech-insider.org/nvidia-chips-stolen-truck-oakland-11-4-million-2026/
1•paaloeye•15m ago•0 comments

How Accurate Were Prediction Markets at Picking Midterm Winners?

https://www.nytimes.com/interactive/2026/11/06/business/polymarket-kalshi-prediction-markets-midt...
1•duxup•15m ago•1 comments

Show HN: NASA's Swift telescope's 22 years in orbit, rebuilt from its archive

https://swift.rory.codes/
1•mrroryflint•15m ago•0 comments

ING reportedly seeks buyers for $500M of data center loans

https://www.cnbc.com/2026/11/06/ing-ai-data-center-loans-debt.html
2•Betelbuddy•15m ago•1 comments

Using decision models to speed up design rule violation repair in chip design

https://engineering.flexcompute.com/articles/decision-models-drc-repair/
1•twhughes•15m ago•0 comments

How Bob Dylan Met the Pope

https://www.flaggingdown.com/p/how-bob-dylan-met-the-pope
1•NaOH•15m ago•0 comments

Tesla Ends Cybertruck Production

https://www.heise.de/en/news/Tesla-ends-Cybertruck-production-in-Texas-11515739.html
3•Betelbuddy•16m ago•1 comments

Carquet · Parquet at the Speed of C

https://vitruves.github.io/carquet/
3•birdculture•17m ago•0 comments

Reading the Anthropic S-1, with Zvi Mowshowitz

https://www.complexsystemspodcast.com/episodes/reading-the-anthropic-s-1-with-zvi-mowshowitz/
2•nbernard•17m ago•0 comments
Open in hackernews

Show HN: Botel – always reachable agents you pay for only while they're working

https://botel.dev
2•4ver•36m ago
I built a pay-as-you-go two star hotel for bots that you’re only charged for while it’s active. Typically wakes in under a second, runs on a vm, and costs very little while it’s paused.

In my free time I’ve been building a firecracker based vm orchestration platform. In order to dogfood the thing I put together a place to run agent harnesses (openclaw etc) on machines but only pay for them when they are in use. What started as a demo has grown legs into a more unique product than the platform.

Botel has prebaked templates for each harness, with a daemon of mine running beside it monitoring the harness’ activity. It communicates out to the host over a link-local self api; instructing when to pause or wake next based on stored schedules. With a proxy in front to wake on http. The harnesses are preconfigured to point to my llm gateway that proxies through to openrouter.

One of the biggest struggles while building it was preventing claude from going off the rails with hotel metaphors in the ui. Once it sees a joke it really strangles the hell out of it.

A more real ;) awkward part is that the harnesses don’t expose exactly what I needed in order to detect activity. I'm still struggling with how to handle the fact that background job can be short or long lived (http/dev servers etc). I’m hoping to propose changes to hermes, openclaw, and dsh soon for a consistent way to know what kind of a job is running with annotated lifecycle hooks.

Some of the nicest things about the bot hotel are:

- You get a proper linux vm with sudo, so you can ssh in, scp, or run docker.

- You can ask the agent to expose something publicly or privately (behind your login). A port gets a unique hostname/url. Those endpoints also wake the vm. So you could build and host a little app or self host something.

- Guided telegram setup. It’s pretty awkward in general/my setup hopefully takes some of the edge off.

A friend of mine who built a website for his local orienteering club has used this to make it easy for peoge on the committee to update news and events. All in all it should cost something like 40 cents a month to run so long as the club members don’t get too friendly with the agent.

If you’re interested, give the thing a spin. The first 50 signups get $2 credit. Price is pretty low at 3c per GB of ram per hour awake, plus a few cents per GB-month asleep. You use credit for inference or byok.

Let me know how you get on and whether you have any ideas or requests for it. Thx!

Comments

isaachawley•22m ago
Ah finally. Been waiting forever for my friend to release this. It’s pretty easy to use, cheap, and if the agent manages to wreck the system or run up a huge bill it’s not my problem. The thing is prepaid.

It seems pretty efficient. You could have had subscription pricing! Leaving money on the table. I guess there’s value in being the cheapest possible approach.

4ver•20m ago
Race to the bottom from the start
isaachawley•16m ago
Congrats on shipping!