Hacker Newsnew | past | comments | ask | show | jobs | submit | jffry's commentslogin

"Workplace violence restraining orders" in California appear to be a type of restraining order that can be filed by an employer on behalf of an employee, to protect said employee from a third party

https://selfhelp.courts.ca.gov/WV-restraining-order


> We've been including product tips in PRs created by Copilot coding agent

If the PR is wholly authored by Copilot I get the spirit of this, although maybe not the best implementation. And "tips" like this that look like an ad for a product _definitely_ feel like an enshittification betrayal of the user, even if it was a genuine recommendation and not a paid advertisement.

In the OP's situation, where where Copilot was summoned to fix some thing within a human-authored PR, irrelevant modification of the PR description to insert unrelated content is specifically egregious. Copilot can easily include the tip in its own comment, so I'm curious why it was decided to edit the description of a PR instead.


Nah, PR text is a completely inappropriate place for a tip to appear. A PR description should describe the contents of the PR, not include unrelated, unsolicited advice. It’d be like submitting a bug fix, and saying “this PR fixes bug X, and also, have you considered using a different linter in this project?” Completely inappropriate.

To be honest, just a user here, it’s only recently (like a week?) you can ask Copilot to edit an existing PR, historically it’s had to open a new one (that merged back to original PR) or it had to make it to begin with, I can see this unintentionally happening as part of this improvement to edit existing PRs

It's unnecessarily splitting hairs.

> interaction data—specifically inputs, outputs, code snippets, and associated context [...] will be used to train and improve our AI models

So using Copilot in a private repo, where lots of that repo will be used as context for Copilot, means GitHub will be using your private repo as training data when they were not before.


No it isn't. Most people don't use Copilot, so this term change won't effect most people. You can reasonably be unhappy about it anyways (or unreasonably still be using Copilot in 2026), but it's still ultra-useful information for them to add to the discussion.

Next step they'll rebrand search as "Copilot Search" or auto enable pull-request AI reviews (unless you hear about it and turn each off) and we'll all be "users".

Boiling the frog with a Venn diagram.


Copilot, or "chat with Copilot" is a button that is available on every page right next to the search bar.

I don't have to be a Copilot user to click on it.

This change is malicious, and it doesn't only affect Copilot users. It affects everyone on the platform!


Again, this collects usage data. If you click the button by accident and don’t interact, they get no data.

So? This feature is available to everyone and you have zero idea how many people actually use it.

If I go to one of your GPL projects and I ask a simple question to find out what this project is about, you will be perfectly "ok" that this interaction (that includes most of the code that is required to answer my dumb the question) will be used for training?

This is not ok.


Nobody in this subthread is saying if it's OK or not. We're just saying that it's very useful to know that this is what they're specifically collecting. Jiminy.

It's automatically enabled for example the other day I did a commit directly on GitHub and AI generated commit popup it had to read the code to work

> Most people don't use Copilot

So why do any of this at all? You're putting a large part of your customer base on edge in order to improve a service that "most people don't use." The erosion of trust this brings doesn't seem like a worthwhile or prudent sacrifice.


You're asking me to explain Microsoft AI strategy? Your guess is as good as mine.

I don't use copilot, but somehow was subscribed... I probably clicked something long ago and it just remained active.

They "gift you" a free standard plan if you have above a certain (non-transparent) level of stars, I don't think you can even disable your "subscription" if you get it for free.

They're only training on interactions with Copilot, not with the full contents of repos that happen to be subscribed to Copilot.

Make it opt-in then.

Isn't this pretty standard, using your interaction data for training and making it opt-out? Claude Code, Codex, Antigravity etc. all do the same. Private repo doesn't make a difference as they have a local copy to work from.

Nope, it's pretty compact. The first image in CERN's photo gallery shows it being loaded into that truck: https://cds.cern.ch/record/2957407?ln=en

Of course, it's compact because it only has to last so long. CERN's press release discusses needing a generator and a cryocooler in the truck for longer trips: https://home.cern/news/press-release/experiments/base-experi...

This older article about the test they did with ordinary protons, indicates the outer frame measures "2.00 meters in length, 0.87 meters in width, and 1.85 meters in height" and comes in under 1000kg https://ep-news.web.cern.ch/content/cerns-base-step-leap-for...


It's a tech demonstrator for a company that turns models into custom silicon for fast inference. In this case llama3.1-8b https://taalas.com/products/


Is this an ASIC? Or FPGA? Or something even more exotic?

I’m guessing it’s some form of ASIC because I can’t imagine crafting the logic of Llama on silicon is a very quick or easy job. Not that doing it on an ASIC is a piece of cake either.


An ASIC is custom silicon, no?

Anyways, I found this article discussing it a bit more: https://www.eetimes.com/taalas-specializes-to-extremes-for-e...

"Taalas is borrowing some ideas from the structured ASICs of the early 2000s to make its hardwired model-specific chips. Structured ASICs used gate arrays and hardened IP blocks, changing only the interconnect layers to adapt the chip to a specific workload. At the time, this was seen as a more cost-effective alternative to a full-custom ASIC that was more performant than an FPGA."

"Taalas changes only two masks to customize a chip for a specific model, but the two masks can change both model weights and dataflow through the chip. On the HC1, the model and its weights are stored on the chip using a mask-ROM-based recall fabric paired with a (programmable) SRAM, which can be used to hold fine-tuned weights and/or the KV cache. Future generations of chips may split the SRAM onto a separate chip, meaning they could be denser than the HC1."


For a sense of how crazy 30 detections per day is: Super-K is a cylinder 41.4m tall and 39.3m in diameter [1] and neutrino flux on Earth is 65 billion per square centimeter per second [2]

The tank's cross sectional area relative to the sun depends on its relative orientation to the sun. We'll ballpark it at somewhere between its circular endcaps (Pi x ((39.3/2)^2) = 1213 square meters) and its curved cylindrical face (which, pointed right at the sun, has a rectangular cross section of 41.4 x 39.3 = 1627 square meters).

So, conservatively, the neutrino flux through Super-K's tank is 1400 m^2 x (100cm/m)^2 x 6.5e10 neutrino/second/(cm^2) x 86400 seconds/day = 7.86e22 neutrinos/day passing through the tank. Of which 30 are detected.

[1] https://en.wikipedia.org/wiki/Super-Kamiokande

[2] https://en.wikipedia.org/wiki/Neutrino (end of intro section, just before History). Wikipedia says only that the "majority" of the 65 billion flux is from the sun, so we might be off by a factor of two-ish in the worst case.



I've done the same thing - making a USB-only printer available on my LAN - following this guide: https://pimylifeup.com/raspberry-pi-print-server/

One nice thing is I can print to the CUPS server even if the printer is off


BTC cannot be split beyond 10^-8


With a soft fork it could easily.


Lightning Network already uses millisatoshis. Of course they can't be settled in sub-satoshi amounts on the main chain, until there's enough interest in a fork to do so.

The fixed supply describes the total sum of units that have been issued, and that are intended to be issued in the future - it doesn't relate to the divisibility of those units.


Thanks , I've learned something new today!


Every time I've looked into doing a DIY NAS in the last few years Topton seems to come up - as far as I can tell it's because they make MiniITX boards with a boatload of SATA ports.


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: