3eb7988a1663

Ok. Are they lying?

The entire post is that Tesla setup $30B in loans that it could potentially use. What salt should I be taking?

  XGBoost’s search optimized accuracy, and afterwards I also compare by area under the curve. Which means the “fourteen of fourteen on AUC” is against a boosting model that was not tuned for that metric. Tuning it for AUC would probably improve it there; I did not measure that.

So, not a fair test?

I also did not get why Xgboost had to count its training time for the inference. You only train once. I guess in some scenario, where someone says, "I need the best model now, you have five minutes on this singular dataset", but I have never been in that situation.

I would feel better if scripts were released, because I am fairly dubious. I take it as a given that a tabular model has been pre-trained on all of the public benchmark datasets, but that is what it is.

The slop was so meandering, I am not sure what is truth or not.

I would push back on that. It used to be less terrible before they made it all SPA. Now there are all sorts of broken-caching tricks that try to do in-place reloads. I frequently see progress bars incrementing while it is attempt to do a request - while that original redraw is attempting to happen, I can frequently launch a new tab with the same link that will finish rendering before the original.

I write Python. While I read the original post and nodded along, nothing changed about how I worked until I started playing with Rust. Seeing a strongly typed language force the constraints and stop stupid errors was when the rationale finally clicked for me. Knowing that this function will never blow-up in an unexpected way changes how you think about problems.

You otherwise get accustomed to some wishy-washy blobs of data that get passed around and find it normal. Maybe you include some ad hoc guard rails here and there, which catches the egregious errors, but there is always some lingering uncertainty. Some string that should have been an int, missing key here, the object which never had the validation check, etc

It is like unit-testing - more-up front work, but I could never go back to a world where I did not get these automated assurances. Sadly, I am a grug-brain which could only feel the lesson from personal experience.

Don't all of "serious" programming languages meet that bar? Java, C#, Go, Python, etc all have enormous test suites. Once you get into the third party, things become much more uneven, but if you can restrict yourself to say the top N packages in a language, those are going to have better than average development practices which makes that plausible.

That's news to me. One of the few real powers of the board is to fire the CEO. You don't need to put the CEO on a PIP first. If the board thinks the CEO could do better, that is all that it takes.

I have animations minimized (whatever the setting is called) on my iphone, and there is still a maddening delay when doing many actions. Either the app has hard-coded some amount of time between presses or the program is so inept that it really takes fractions of a second to do basic activities.

I think it is worth noting that older data centers were not materially moving the electrical grid. That Utah data center is planning to use as much power as the entire state.

Regardless, it still comes down to: things are bad and unaffordable. The powers that be keep claiming that AI is about to take our jobs and end the world. Why wouldn't people be mad?

I have no sense of gwen's readership numbers, but I find it unlikely that one rather erudite blog has anything to do with global nicotine consumption. People have been smoking for centuries. The vaping industry had guerilla marketing campaigns to target college kids.

No rain drop thinks it is responsible for the flood, but I am hard-pressed to imagine a blog article greatly swayed hearts and minds.

Global fossil fuel usage fell last year. We are on a trajectory to massively cut out fossil fuels and replace them with renewables. This was before global petroleum movement was blocked through the Strait. Check out the juicy details in the Ember 2026 report[0].

Some highlights:

  Driven by record solar growth, low-carbon power generation increased by 887 TWh in 2025, outpacing electricity demand growth of 849 TWh. Solar power alone met 75% of the net increase in electricity demand. Together with wind, the two sources met almost all (99%) demand growth. For the first time since the Covid-19 pandemic in 2020, and only the fifth time this century, fossil generation did not rise, recording a small fall of 38 TWh (-0.2%).

  The global fall in fossil generation was driven by a historic reversal in fossil trends in China and India, the largest and third-largest fossil power countries globally. 2025 was the first year this century when fossil generation fell in both countries. In China, it fell by 56 TWh (-0.9%), marking the first decline since 2015.

  For the first time in 100 years, renewables (33.8%, 10,730 TWh) overtook coal power (33.0%, 10,476 TWh) in the global electricity mix as continued rapid growth in solar and wind pushed the share of renewables above a third of global generation. Coal power dropped 63 TWh (-0.6%) in 2025, marking the first fall since the Covid-19 pandemic in 2020. Combined with continued electricity demand growth, this meant coal fell below a third of global generation for the first time in history.

[0] https://ember-energy.org/latest-insights/global-electricity-...

No, I get that. The pondering was before I knew it was a recording of actions. Once you know that, it is a fully self-contained box with external nudges from the players which get inserted into the simulation.

I suppose that makes sense for a deterministic engine. That way you have an ordered steam of actions with only one way to execute them.

I was wondering how you could possibly sync up something like a Carrier that has multiple fighters.

You burn a barrel of oil once. A solar panel lasts for 20+ years. After you have installed it, if China announces they will no longer sell new ones - you have two decades of lead time to start building your own domestically.

Or that Germany shares electricity with the EU, so that everyone is getting subjected to those pains with fuel sources being cut.

One thing that I really want to know - the better models from today vs a year ago - what has changed. They have already pre-trained on all available public data. Scooping up the last percentage of archaic texts which were never digitized is not going to move the needle.

Is it just that the providers are generating tons of synthetic datasets on coding tasks so that the models get more exposure to the right thing to do? Every time someone points out an LLM stupidity they add some training data to patch over the weakness (trivial to generate "there are two 'l's in llama")?

Do replay files require consistency or can you spawn new units from the ether as required? I think you could get a system that would get the broad strokes unit placement correct, but having a shot-for-shot perfect replication seems impossible.

Who is offering the refusal vectors or instructions on how to generate them?

Formal agreements matter unless the counter-party is the US who will gleefully alter the deal the second it becomes inconvenient.

Would that hold up in court if the person were willing to challenge it? That seems a clear cut violation of that person's speech, but I guess anything for "national security".

Indeed this is not the article I originally read either, but very similar breakdown of the problem.

The article did mention suction sails for the Maersk Tankers, but it does seem like the bulk of the article was about the rotating tubes.

If anyone else was confused why those pictures were of giant tubes and not canvas sails, you are not alone. Apparently those are some version of "Suction Sails". The vendor's website has a graphic that explains it, though it was fighting my ability to scroll.

https://bound4blue.com/esail/

Significantly larger, yes.

At the conclusion, he said he spent two years on this! That's a thesis.

I am failing to find the article, but some early online poker systems used the server time as the seed coupled with a weak PRNG. With the information of the hole cards + community cards, after a few hands, could quickly determine exactly what seed was being used and perfectly predict everyone's cards.

I can think of at least two people in my career who were, ahem, load bearing for the company. They did all of the BS, non-sexy work required to keep the place running. Never wasted time on vanity projects, just showed up day after day doing the actual necessary things.

They were criminally under appreciated by management. Their career trajectory significantly worse than those who engaged in the highly visible projects.

Skyrim came out in 2011 and is at this moment, the 55th highest player count on Steam. Just playing top tier single player games over the past twenty years could keep you incredibly busy.

That's exactly how it was revealed that a high quality image intentionally leaked by Trump came from a satellite and not an overhead plane.

Press article: https://www.livescience.com/secret-classified-satellite-trum...

Blog write up: https://sattrackcam.blogspot.com/2019/09/image-from-trump-tw...

From a purely efficiency standpoint, yes, but practically, electricity->heat is a lot cheaper to install. Plumbing is messy, and bolting on a water circulation system to a roof is harder than delivering more joules over the wires.