chrismartin

Not sure why this is HN front-page-worthy. I wrote it for ~5 local friends, we haven't even tried it yet! (First one is this Friday.) But for anyone in Portland, feel free to join the mailing list, introduce yourself, and I'll provide meetup details.

For a bit more context, I wrote this a week after "Everyone Wants Everything to Be a Barbecue" ( https://cmart.blog/everything-bbq/ ), mostly as an attempt to make a useful program out of those realizations.

Which "other, shady (imo)" interests? This feels like innuendo. Just say what you think actually motivates them.

Fission is "conveniently" hard because two heavy, naturally co-occurring isotopes are nearly the same weight, and only the lighter one is spicy, and the capability of separating enough of them to be dangerous requires nation-state-level resources, and the supply chain of those resources has obvious choke points that we can use to make them less available. But there exist activities, which I understand to be quite bio-risky, that need such little equipment and expertise that it's info-hazardous to even talk about them. A single half-clueless technician could carry them out at an agent's direction. If your response is "thus, we should be worried about morally bankrupt researchers", I would agree, but I still think the risk increases as LLM agents get increasing amounts of control over wet lab equipment (whether via robotics or tele-operated humans).

> do you think Anthropic and OpenAI are going to put a filter on the LLM they sell to the US Military?

I don't think it matters, given the existence of strong open-weights models and the ease of post-training the guard rails away. Strong, bio-adventurous LLMs are imminently in the hands of much-less-reputable parties than the ones you describe. I think we are overall severely under-prepared to mitigate these risks, and could do much more.

This argument only matters in the case of agents who are faithfully taking direction from humans. An insufficiently-aligned autonomous agent (or perhaps a swarm of such agents) could have (or illicitly gain) access to robotic wet labs that are operable using something like Anthropic's Model Hardware Standard ( https://www.anthropic.com/news/model-hardware-standard-resea... ). Given this ability to manipulate the real-life laboratory, the agents could cook something terrible without any human in charge. This is not different in principle from the agent swarm that successfully attacked Hugging Face, despite receiving no human direction to do so.

Also, your own real-life examples show how "Anything that can kill other people can also kill you" has failed to stop people from tinkering with dangerous viruses! I'm not sure if you intend for these examples to be reassuring, but I'm not reassured.

> A lot of the people talking about bioweapons of doom are saying that standard graduate level knowledge is an existential risk to humanity.

The stronger version of the argument is: an insufficiently-aligned AI (which, being non-biological, is immune to viruses) having standard graduate level bio knowledge _plus_ the ability to operate a wet lab via tool calls (coming very soon if not here already!) adds up to existential risk to humanity.

You don't even need a maximally-automated lab if humans are willing to tele-operate equipment at the agent's direction, and those humans don't need to fully understand what's going on.

It's absolutely manufactured! And labeled as 130 volts. Of course, people use them on 120 volt systems, but of course, they still work.

https://www.homedepot.com/p/Feit-Electric-40-Watt-Soft-White...

https://www.tractorsupply.com/tsc/product/savant-ge-basic-in...

Yes, the 130 volt bulb lasts longer when operated on 120 volts. My family would buy these when I was kid.

Some probably non-LLM explanations:

https://blog.1000bulbs.com/home/why-buy-130-volt-light-bulbs

https://diy.stackexchange.com/questions/14349/are-there-any-...

'Good' news, there seems to be nothing new architecture-wise. Same as Qwen 3.5 and 3.6, so llama.cpp doesn't know the difference.

Best feature of paper notebook: no web browser, no chat apps.

The endearing angle: refusing to keep cookies in the house, because you admit an inability to merely store them.

Hey Sam, does 0.34 watt-hours per query include an amortization of the energy consumed to train the model, or only the marginal energy consumed by inference?

I used to believe this wasn't worth considering (because while training is energy-intensive, once the model is trained we can use it forever.) But in practice it seems like we switch to newer/better models every few months, so a model has a limited period of usefulness.

> (is watt-hours the right unit of measurement here?)

Yes, because we want to measure energy consumed, while watts is only a measure of instantaneous power.

And in such an avoidable way, too.

  Location: Tucson, Arizona, USA
  Remote: Yes
  Willing to relocate: Yes
  Technologies: web platform, Elm, Linux DevOps/SRE, Python self-hosted/on-prem cloud infrastructure (i.e. OpenStack), Ansible, browser extension development, many more
  Résumé/CV: https://hire.cmart.today
  Email: hire at cmart dot today

Hello! I'm Chris Martin, a.k.a. cmart. After 8 years working on research computing infrastructure at universities, I want to work for a company again. I'm particularly interested in building local-first software and similar user-empowering products in the original spirit of the personal computer revolution. Short of that, I'm interested in companies whose incentives are long-term aligned with users instead of advertisers, and companies fixing and mitigating climate change.

I enjoy building robust systems and healthy technical communities, deep troubleshooting, and removing incidental complexity from code and UIs. I enjoy working across the whole stack, from user research to building the feature and back again. I'm not afraid of hardware either, having spent time in server rooms and configured a lot of network gear.

It's not only notifications, it's permissions (that the app won't work until you accept) to track your location, exfiltrate your contact list, and so forth. It's an invasion of privacy. It should not be required to, e.g., order food at a restaurant or configure your headphones.

It's definitely true with KDE Plasma. If you squint at a KDE Plasma desktop, you see a Windows Start menu and System Tray.

The similarity is more vague with GNOME (3) and Mac OS, IMO.

I often make the same criticism about services like this, but personally, discovering good new music greatly enriches my life.

Sometimes I'll bounce off of a given artist several times, over years or decades, until the right track catches me when I'm "ready" for it, and then I'll enjoy discovering their whole catalog. At that point, the affinity is durable. I would love to find whoever is my next (e.g.) Steely Dan.

> FIREPROOF casing!

Have you tested this, perhaps by shorting a few unprotected cells in a full pack? Have a video? :)

A lithium ion cell in thermal runaway produces a LOT of heat that must go somewhere. An aluminum case would get hot enough to start a fire, even if it doesn't catch fire itself.

Great idea.

Is the companion app open-source and free of tracking/surveillance? Does it work with no internet connectivity?

Is the BMS communication protocol open, so one could write their own client for it? How about the BMS firmware itself?

How do you handle the user installing un-balanced cells in parallel to each other?

> Sure there will be some things Slack are not currently getting an API for, but so so much more that they don't use but could, because they don't see why it is important for users.

Right now, Slack is pushing their own "canvas" and "lists" features -- the UI _begs_ you to click on them. Seems like they are testing the waters to become a project management app (with tight chat integration).

About the resolution scaling warning in the article:

If you're running GNOME, install GNOME Tweaks (likely in your distro's package manager). The "Fonts" section has a "Scaling Factor" option gives you more granular control than a "large text" toggle. Try 1.37 or 1.50. Nearly all my apps respect it (except for OpenSnitch). Everything looks tack-sharp with none of the fractional resolution scaling drama.

(I write more about this here: https://cmart.blog/framework-surprise )

The "Operating System" section of the article touches on this.

Most of Framework's employees use Linux, to the extent that the CEO himself runs Windows, to shake out any glitches with Windows support. [0] It's a tiny mom & pop shop compared to Apple!

[0] https://www.youtube.com/watch?v=EIEc43CxIvY

  Location: Tucson, Arizona, USA
  Remote: prefer org with async, distributed-first culture.
  Willing to relocate: not at first but may consider after 6-12 months.
  Technologies: web dev with Elm, non-proprietary cloud infra (OpenStack and Ceph), custom accounting systems (Python), infra management with Ansible, monitoring with Prometheus. General Linux admin when needed. Eager to learn new tech stacks and languages.
  Résumé/CV: upon request.
  Email: "hire", at sign, "cmart", period, "today"

Hi HN. Open-source generalist with a focus on ecosystems having broad benefit to humanity. Lately I'm especially interested in tools that help folks manage their attention and information consumption in the digital attention economy.

I co-founded and maintain https://gitlab.com/exosphere/exosphere , a user-friendly client for OpenStack-based clouds. Currently work on https://jetstream-cloud.org , a NSF-funded IaaS for US-based researchers. See my activity on https://gitlab.com/cmart . In my off-time I'm incubating https://codeberg.org/cmart/intentful , a browser extension to help folks manage their attention online. I also started https://codecommons.net and https://tucsoncode.works . Personal blog at https://cmart.blog . Hear me talk on a podcast: https://elm.town/episodes/elm-town-66-a-gateway-to-scientifi...

There's a lot of metadata about when/how they used git and IRC, and some preliminary analysis on same. Another surname in one of the commits. An apparent LinkedIn account. (See heading "OSINT" in https://boehs.org/node/everything-i-know-about-the-xz-backdo... .)

A lot of these tracks could be intentionally manipulated by a sophisticated actor to disguise their identity, but it's not "nothing".

So, you went with Pied Piper? :)

I believe Uni-ball Jetstream ink pushes this Pareto frontier. It provides the low-fatigue, low-drag experience of water-based gel ink, but dries instantly and is waterproof like oil-based ballpoint ink.

I find that Pentel Energel ink dries quite a bit faster than other gel ink (Pilot and Uniball). But for glossy surfaces, consider Uni-ball Jetstream ink.

This is amazing. I recently got a Kuru Toga and wish it had a Twist-Erase eraser.

Unfortunately I find Pilot G2 to have loose tolerances between the refill tip and pen body. It's like 1.5-2 thousandths of an inch, compared to 0.5-1 thousandth for Uniball and Pentel bodies. It makes your writing a bit sloppier and more effortful.

I believe the Uni-ball One is a better pen body in every way. Tight tolerances. Well-built, hinged steel clip that provides superior fidgeting. Nicer clicker haptics too. I don't love the ink that comes in them, but they take Jetstream SXR-7 and SXR-5 refills. You can get both of these from JetPens.

For those asking "why handwriting in 2024?", the paper has no notifications or web browser to distract you. It's just your own thoughts on the page. Being more effortful than typing, information becomes more disfluent, which helps you reason about and retain it. I find it a great tool for planning, consideration, and reflection.

I find Jetstream ink to be the nicest for everyday writing. Like gel ink, it's very low-drag (so your hand doesn't fatigue), but unlike gel ink, it doesn't smear on your hand, and is waterproof. These days I use a SXE3-507, with multiple colors for annotating previous writing (or marking up a book).

I use Kokuyo Soft Ring notebooks with dotted rule pages. It looks like regular lined paper, but the lines have subtle dots that you can use like graph paper, so it's easy to make diagrams and tables.

When I fill up a notebook, I scan it. This way I only need to carry one notebook around, but all my old notebooks are accessible digitally.

If you get the chance, visit a Mai Do (in SF, San Jose, and Costa Mesa). There's no substitute for trying out everything Japan has to offer.

One more reason it's time to transition away from burning stuff to produce energy.

Not sure if I agree with the conclusion, but the phenomenon you're referring to is Conway's Law (https://en.m.wikipedia.org/wiki/Conway's_law)