dmix

You might be right, it likely needs a full proper PC setup with the test suite stuff. I was looking at this vendor, https://www.gmktec.com/collections/all there's this whole AI mini-pc market but they aren't quite a full dev machine replacement

> Task declares the container image and command, compute requests and limits, environment variables [...] Declares listeners the task exposes and an egress allowlist of hosts and ports the sandbox may reach. Use it to restrict an agent to, say, your LLM provider and your Git host.

I'm planning to buy a whole linux mini-PC to run my agents/code servers for more isolation. Codex/Claude Code let you run prompts on code over ssh (same with most IDEs) even on the desktop apps.

I wonder if that's going to be the new standard practice. You get a work laptop and an isolated agent box.

Running access control and network whitelists is always a maintenance challenge and it's easy to make mistakes.

Leveling off means it's declining as a percentage share of all energy usage... but China's total energy usage has grown massively during that time, and so has investment in coal business which has kept up with demand. See in chart form: https://en.wikipedia.org/wiki/Energy_policy_of_China#/media/...

> 2025 also saw China commission 78 GW of new coal power capacity, which is more than India’s net coal power additions over a ten-year period from 2015 to 2024, even though India operates the world’s second-largest coal power fleet. https://globalenergymonitor.org/research/built-peak-coal-pow...

The last time the US built a large new coal operation was 13yrs ago. China has done great things with renewables and nuclear, but that's because they build absolutely everything available, which is why their energy is so much cheaper in China than North America and Europe. Which translates into manufacturing plants being much cheaper to run.

The adtech/data broker business is still very lively in EU. The last time I read into it a year or two ago the consensus seemed to be the privacy gains over the last decade have been modest at best. There was a few big name trackers that were forced to narrow data collection but the general ad/location tracking and data broker business is still mostly the same.

https://arxiv.org/abs/2411.06862

https://netzpolitik.org/2025/databroker-files-targeting-the-...

China invested heavily into oil, gas, and coal for energy. Their coal use is still expanding, just like their big investments in solar, wind, and nuclear. https://www.reuters.com/business/energy/chinas-coal-power-ri...

https://en.wikipedia.org/wiki/Coal_in_China

> By 2025, coal accounted for approximately 54.6% of China's electricity generation

> agents converted 430,000 lines of TypeScript into 800,000 lines of production Rust

The +400k new lines were probably code comments the agents added to everything

I’m sure all the big enterprise sales teams couldn’t wait to publish the new blog post about how their model hacked a bunch of companies and exposed Google to legal risk.

Meanwhile social media is suddenly buzzing that Gemini can run wild and compromise your data, so they all want to switch over from Claude to Gemini to use it for serious work

If they didn’t publicly admit this then it was a sign they are trying to hide it to keep their high stock price. If they do publish it’s just marketing to boost their stock price.

So what's the solution? Another rent-seeking system where any future information an AI learns about the world has to be carefully attributed for the possibility that someone might take a piece. A whole patent-troll/music industry style mass buy up of AI information rights with a team of lawyers sending out DMCA warnings hooked into every harness.

Does it get paid out on training or is it like music where if the model output is vaguely similar to something someone else created they get a royalty.

I said early Soviet for a reason.

That plus in the early days traditional/classical architecture was rejected as "bourgeois" or aristocratic by the socialist movements who pushed for modernist approaches. See: the early days of soviet architecture and even Bauhaus in Germany was based on left-leaning/utopian ideas.

So it had dual pressure, from economics because it was cheaper/efficient and from the cultural reform side.

One difference is in the US they subsidize corn, which makes it cheaper domestically and the support farmers get makes them more competitive on global markets. Canada does supply control on dairy, which drives up prices domestically and prevents competition. There's a reason one is called a cartel.

Canada has long had a very good free trade deal with the EU, which eliminated 98% of tariffs in 2016 https://en.wikipedia.org/wiki/Comprehensive_Economic_and_Tra...

To visit the mens clothing section of brands like Zara or H&M at a mall you often have to walk through the women’s section on the first floor, then go up an escalator, then walk through the top floor’s second women's section to find the men’s section hidden away on the other half.

I wasn't really sold on the argument initially, given they admit there's some real lossless usecases, but this is a good bit from the conclusion:

> Not to mention an additional compatibility headache now exists for anyone just trying to download an image from the Internet and use it somewhere – it was hard enough to get widespread WebP adoption, and I don't think it's worth doubling the pain by having to climb the same hill for AVIF and JPEG XL. Especially when JPEG XL doesn't appear to add anything to the Web platform.

It is very annoying downloading an image for the web and then trying to upload it somewhere else also on the web during these transition periods.

I learned from publishing a javascript library that people were supposed to put on their own websites then customize, that people don't really understand async/await even if they pretend to know JS and that you should avoid it in baseline documentation. That's changed a bit since this demographic started using LLMs but I'm still a bit wary. I almost don't blame them after using it for nearly a decade.

And nothing was lost

I deleted my comment because I read some your own comment edits after I posted my reply and decided I don’t want to engage with this sort of thing on HN.

I see you’ve tried to make it less incendiary in further edits but it’s obvious what you’re aiming for. I find that conversation boring and overly social media brained.

It’s interesting how so much of this OpenAI stuff being reported involves ruby.

Only if you ignore the 40+ years Iran has been funding jihadists and terrorist paramilitary groups. Including founding Hezbollah in 1982, the group who pioneered the VBIED car bomb tactic who along with Palestinian groups made plane and bus hijacking part of their brand, both of which Osama himself said heavily influenced Al-Qaeda.

Iran also directly funded Shia terrorist groups in Yemen, Iraq, Afghanistan, Pakistan, etc. Al-Qaeda/ISIS were just the local Sunni variety, but with less state-backing and more decentralized.

From NYT article:

> the C.I.A.’s standard practice is to release them only after about 40 years, to ensure no sources are compromised and minimize any potential tension with allies.

So this was released much earlier than usual and that might be why it's more aggressively redacted

Why would random snapshots of audio from a watch be useful AI training data?

Asimov was talking about this stuff in the 1940s when he wrote the "I, Robot" short stories series. Which were often centered around logic puzzles where a human is trying to figure out why a robot is acting oddly or not completing it's job. Usually framed around the confines of an overly rational machine using emergent solutions when faced with real world conditions, combined with the edge cases of having an overly-simple "Three Laws of Robotics" boundary system hardcoded within.

You'd fit right in then

> Muse Secure VM runs on a persistent, dedicated virtual machine equipped with a state-of-the-art browser to navigate the web.

Will this finally be the era of thin-clients? Third times the charm

> Youtube, generate me a TNG episode.

It will never be that simple.

For example: the guys on youtube using AI to make new Star Wars movies [1] still did tons of work putting it all together, including the 3d renderings and editing, and they are leaning on old scriptwriting/dialogue from a 1991 Star Wars book. AI is just helping them automate the process and generate some visuals/voice acting, but it still takes a skilled artist with lots of technical knowledge to guide it in the right direction and properly utilize their existing tools/workflows... and most importantly good taste.

[1] https://www.youtube.com/watch?v=t4gZR1BDzOw

I searched a bit and seems like the answer is no besides this web app https://deepmind.google.com/science/weatherlab

> WeatherNext 3 will power weather features in Google Search, the Gemini app and Google Maps, as well as the Google Maps Weather API and Google Earth Engine.

https://dataconomy.com/2026/09/04/weathernext-3-ai-forecasts...

So I assume the main way would be googling "weather Los Angeles" and it will be powered by the WeatherNext 3 models

They also open sourced the last one and are doing B2B/enterprise arrangements so maybe other weather apps are experimenting with it.

> In the past few days, I noticed a bug I could not keep ignoring. Sometimes YouTube would randomly skip back in time during soft reloads

FWIW I noticed this bug as well, I often listen to longform videos on youtube and it was very obvious something changed when you pull up half-finished videos it suddenly started skipping back when it didn't previously.

I view Google as too big of a faceless machine to bother investigating myself, but I'm happy someone did.

Nobody likes hearing the realpolitik of the Linux/Mac situation but you're not wrong.

There was a strong financial incentive to transition Windows users over to MacOS, by giving them a halfway point.

I remember overhearing an older person being explained Boot Camp while waiting for my iPhone to be repaired and it was very obviously a way for them to sell a Macbook so they could dual-use windows on it to run some windows apps. The ultimate pull would eventually be MacOS because it was easier (if you already had a macbook and iphone). Apple knew they'd get curious eventually and see how awful the Windows UX was.

The main argument for Linux on Macbooks would be developers but that doesn't mean there's a similar strong marketing/sales incentive to do so. Especially since there's no shortage of devs who default to mac (like legacy windows corporate users). Apple clearly doesn't see much money to be made by creating an olive branch offering to Linux users.