arjie

They blocked geothermal drilling in the desert. After multiple years of blocking it with use of environmental law they finally managed to preserve it as a conservation area for sustainable tourism:

> Organizers of the Burning Man festival in northwest Nevada have reached a settlement with a geothermal energy developer that cancels an exploratory drilling project that some feared would ruin the ambiance of the counter-culture event.

https://www.eenews.net/articles/burning-man-to-buy-back-geot...

I enjoy forums, so I prioritize spam filtering to preserve signal to noise. Others may prefer to read every cialis ad. I think these two extremes and everyone in between can coexist on the same forum, but in order to do so it helps to have adequately tuned filters customizable on a personal level. In more detail if you care to: https://wiki.roshangeorge.dev/w/Overmod#Philosophy

I was primarily motivated to do this by repetitive commentary which is indistinguishable from bot content, and I continue to be motivated by this.

Growing up in India, this was a huge issue back then. Sand miners were sort of like your alcohol runners. Big mob operations with connections into a bent police force at a high level.

One of those things which needs a strong state to protect the commons. Weak state power is easily outmatched by individual agents of the state conspiring with the mob.

The site is fairly politically polarized. However it is not a very large subset participating. Strategically blocking users using a Chrome extension can help you maintain epistemic hygiene. The comments section I see is much sparser and much less prone (though not immune) to human impersonation of robots.

That seems like process error rate. Inspection, as far as I know, samples output and checks processes. But inspected processes also have error rates. That’s how it goes. So the pack you got probably was not “inspected” in the sense that no human laid critical eyes on it and its contents till you.

Funny. Two of OP’s events are prior to the DOGE cuts so they could not be causative.

But these panics are so amusing. Pete’s breaking the trains. Hegseth (was it him?) is crashing the planes. RFK’s eating the infected meat. The Haitians are eating the cats and dogs. We need a prediction market for the next panic.

Huh, what you’re describing is what I assumed it was. I thought it was secured against the asset itself. I can see why they hesitate somewhat. In the US, this is usually a covert bailout for a vote bank. Grant loans for their machines and then the state takes possession and discharges the loan and you have a bailout in disguise.

But you’re right. Everything is in execution and for critical manufacturing it makes sense to pay the fraud price to boost it.

It’s someone’s blog post before a calculus exam. I think it’s fun to do this stuff. On your own site you can write whatever you want.

See, when I was a kid I found something I thought was great and inventive only to find that it was not only long trodden mathematics it was famous long trodden mathematics. I named it after myself for humorous value, the distinction between others who dreamed of finding some novel structure and me being solely that I did not know the famous results and conjectures in the space.

https://wiki.roshangeorge.dev/w/Roshan%27s_Conjecture

I think this is quite entertaining.

Many ancient paradoxes are not really paradoxes. Zeno's ones are resolved today with infinite series.

But this is a real one. Is it possible to describe an arbitrary real number? Almost all reals are not describable. But you cannot find a single such number.

The 'paradox' is that the search itself is self-failing - a broken strategy. Of course, now we have the language of sets and functions between them and cardinalities and we resolve this for us in a way that is meaningful. But still now you know the 'existence' of this thing? Can't be described.

It's interesting because of the property of creating with finite words universes of infiniteness.

It's really a pity that we missed the boat on large nuclear reactors. So much more power would enable so many use-cases and cheap electricity would have enabled us to have much less gas heating. Massive damage to the climate caused by misguided regulation. I tried to write to deregulation.gov on reducing the fund allocation required for fission reactor cleanup but c'est la vie. Death by a thousand cuts and then a few big rocks, I suppose. We lost half a century, but so it goes. Now we should be able to do large-scale solar and wind, and if it weren't for the Burning Man people we could have geothermal too, but we'll get there.

The cheapest card that will run this model very well is a ln unlocked CMP 170HX. But you can run it on a 3090. I run it on an old spare A6000 Ampere. I think I wouldn’t use anything lower than 60 tok/s though, which you can get with MTP etc. I just use a full vllm stack but some people see a lot of speed with ninfer (there are non 5090 ports).

The large RAM Macs are unusable for inference of dense models as of now. Token generation is too slow.

Well TIL. Great tip. I only ever hit that key combination when I look around Finder and find myself unable to go up one level easily from the icon bar - an action I do constantly.

That doesn’t sound right. Silicon Valley is clearly responding to massive demand for AI services. The reason Nvidia has so much money is that people are paying them for something only they can make. It’s not like they’re issuing stock into a meme stock spike or something. They’re selling hardware that companies are demanding in large quantities because those companies need the hardware to provide services that people and companies want so badly that every marginal hardware dollar spent earns a large number of service dollars.

This is probably the most demand-mediated response of all time. The natural form of that is that the guy who we give lots of money to for his stuff is going to have lots to spend.

This doesn’t strike me as a problem. There are sufficient entrepreneurs that market need common to a total 20% of the population will be met. In the USA that 70 million is a sufficiently large number. After all, that’s $700 million of MRR at $10/month.

The things limiting provision of services are physical limits like the audience being widely dispersed or just characteristics like not all sharing the same needs and therefore not actually being addressable.

You obviously are well aware that addressability does not track with income category automatically but it does for many businesses. E.g. payday loans, rent to own, check cashing, dollar stores.

Since a large section of this group is in use of government credits, at the least businesses will evolve to meet the use of those. EBT grocers, Medicaid-focused care, etc.

The majority of the effort will be spent where the greatest economic opportunity is, but entrepreneurs will match their ability to address the market to their audience. It’s not possible for a dollar store to sell GPUs to Baseten. They don’t have the capability to do something useful there.

The one UI invention many web UIs now add is the universal command search bar. I’m most familiar from it from Jetbrains IDEs where it’s been for over a decade and it is a remarkable upgrade on browsing menus. With the KeyPromoter extension on I even learned the keyboard shortcut over time. Good UI pattern and now this universal command search is everywhere: Cloudflare, Mercury, etc.

Love it.

Great library. Really enjoy using it. Agents work really well with the kind of site you build with it and it’s very easy to build. Thanks for the upgrade skills.

That was surprisingly nice. I get to see the far distance on the map. Something has to be done about turns, though. They’re far too whirly for me. Good demo!

Haha, that’s incredible. I saw the MAC listing afterwards too and definitely piqued my curiosity. Terrific stuff :)

All this is true but I prefer writing AI assisted code and I doubt any maintainer can distinguish me from full spectrum bot. It’s just not productive to either of us.

It's a pity about the web, because it's becoming less accessible over time. Regardless, I (and some others) archive many of the things we browse to, so here you go:

https://amber.agentic.church/web/lists.ffmpeg.org/2026082807...

Apparently, not the fonts they use for icons.

Classic problem. I recall spending a long time trying to get Freespace 2 SCP working on the Via Unichrome onboard GPU I had decades ago. Who could possibly have cared for that? The intersection of the two sets was probably precisely one person.

By the way, interesting person to have 512 IPv4s (and two /44s? Aren't those enormous?). Root-level http intentionally user/pass gated on your site? I was curious to see.

Originally, yes. Nowadays, it's trivial. Agents can handle this flawlessly.

Certainly, but there was motivation to do that in the past. My patch would probably line up with the rest of the code in the direction that they desired and the cost of fork maintenance would have pushed me to upstream in a way acceptable to maintainers.

Thank you for responding. Seems like a great choice overall.

This makes sense. Apart from the original thing, I no longer upstream anything. It just takes comparatively more effort than maintaining a private fork with my own idiosyncratic fix. This must mean that even small fixes that non-contributors would make in the past are just happening off the books, so to speak. e.g. I upstreamed a tiny fix to Thrift codegen because some function could be much faster. These days I wouldn't bother. I'd just fork and leave upstream to be upstream.

I suspect many people are like me since it feels like a very normie position to take. That means that contributors are even more likely to be useful because both the drive-by genuine contributors have chosen otherwise and these contributors have increased.

It must be quite painful to be an open source maintainer right now. One wonders what to make of such projects in a future where a feature-list is a sufficient prompt.

At first, I thought that the fraction of people who want this kind of network of peers would be rather small. I thought that almost all people want extensive restrictions on user agents and controlling how the data is revealed to the user. But Bluesky actually does not restrict AI scraping and consumption of data. You're just yet another consumer, subject to rate-limiting etc. whether you are browser or headless agent. I have to say that I am somewhat (pleasantly) surprised that this approach is accepted by Bluesky considering the community is more anti-AI than most. The fact that you're allowed to consume the feed is amazing. Good stuff!

Sorry I missed your comment, Peter. I don't know that it's worth very much to describe. It's not sophisticated. Each component just has an attribute named data-ai-context and content that describes it. This needs no LLM. e.g. charts go into canvas, so their data-ai-context has a textual summary of the content (same as is fed to the charting library); buttons describe what they'd do. Then when you click the "Copy to AI", a piece of JS hits a backend to sign a 15 min token for the same user (as if they'd logged in but with a short token) and pulls all the data-ai-context attributes and gives you a prompt in your clipboard describing the site organization.

It just so happens that doing that tiny first bit allows even stupid models to navigate the site. After that, advanced models also sometimes query for the data-ai-context and they can mimic the Copy To AI to not have to nav the site.

e.g. content "Job Failure Rate: failed sync jobs as a share of jobs run, per connector per day, over 7 days. Day X: facebook=8 failed / 862 jobs; ga4=0 failed / 196 jobs;"

It's really just an alt text element and bizarrely worked the first time I tried it with old Opus 4.5 and I've kept it. Perhaps the biggest use is giving the agent the short-term token.

Wait. I remember reading about this. Doesn’t SBA 7a cover your use case? I’m sure there’s an SBA loan category for you. It’s backed by the machine. Recently they changed the valuation of things or something.

Yeah, I particularly like London’s implementation (I moved away just short of a decade ago - might have changed since) where if you just use the same card you don’t need a pass because there are daily/weekly/monthly maximums and you travel for free afterwards. Very good system. But it is slower that Suica/Pasmo.

That also happens with Clipper Card in the Bay Area. Happened to me a couple of weeks ago on BART. I wonder if all transit cards need to be loaded with money when used on iPhone. I remember having a Clipper Card with autoload and that would work just fine without balance.