There are some weird mistakes too, like for San Antonio hot spring in NM the coordinates are right but the county is wrong. Seems like that should be an easy thing to automate using GIS.
scratch was the trigger for me just as you say. I was turning 11 when we used it at school (28 now). A year or so later I was reading Java for Dummies and making little tools on the TI-84. Then learning C++ and python on a raspberry pi 1B. I’m grateful to that computer lab teacher who thought scratch was worth teaching.
On the last Dwarkesh podcast with 3blue1brown, one of them mentioned that frontier models are now able to work through a whole proof in natural language, just like a human mathematician would. But when they first solved IMO problems in 2024, they relied more on Lean to catch hallucinations.
Luckily: Update from announcer is that trains can start again at 12:25 AM and they reduced our delay by 30 minutes. But there’s still a huge line of riders at the DB service desk.
I’m sitting in an ICE in Munich that was supposed to leave a few minutes before I saw this story on HN. First the conductor announced a 30 minute delay because the radio wasn’t working, and then they bumped it to 2 hours. They didn’t say it was a systemwide problem.
> Early 2026: OpenBrain continues to deploy the iteratively improving Agent-1 internally for AI R&D. Overall, they are making algorithmic progress 50% faster than they would without AI assistants—and more importantly, faster than their competitors.
> you could think of Agent-1 as a scatterbrained employee who thrives under careful management
According to this document, 1 of the 18 Anthropic staff surveyed even said the model could completely replace an entry level researcher.
According to the article [0] that's been making the rounds, NASA didn't make any changes to Artemis 2's heat shield after getting data from 1's re-entry. NASA did change the trajectory for 2, and they made the compound "less permeable" but that change was made before 1 flew.
ChatGPT came out a little over 3 years ago. After 5-10 more years of similar progress I doubt any humans will be required to clean up the messes created by today’s agents.
That’s funny. When I was little I found “format” in my mp3 player’s settings. Thought it would customize the UI or something, but instead I ended up with no music for the rest of the road trip.
Braid also has the stars which are so well hidden that I can’t imagine anyone finding them without a walkthrough (though some people obviously did in order to make the walkthroughs).
The Witness is different, it really does teach you everything you need to 100% it. I cheated on the ship puzzle but it’s totally possible to figure out.
We know angle EBD equals BAC, since the sum of triangle ABC's interior angles is 180 degrees and the sum of the 3 angles at B are also 180 degrees. We also know angle DEB is 90 degrees since DE was constructed to be perpendicular to CB. Finally, D was placed at a distance c from B. The two triangles have the same angles and the same side lengths opposite the right angles, so they must be congruent.
This has precedent in the US, like when the government nationalized failing freight railroads and merged them into Conrail. But after the more recent bank and auto bailouts I wouldn't expect to see this happen again. The shareholders would really prefer to have money thrown at them but also keep their stake.
> the gaps are reduced to zero and the curve crosses through each and every point within its build envelope
I think that's oversimplifying an important point. If you build a Hilbert curve in a 1x1 square, the vertices of the curve always have rational coordinates. So all points on its line segments must always have at least one rational coordinate.
There's no way it can cross through every point in a square region of R^2.
A better way to say it might be "the gaps are reduced towards zero and the curve will pass arbitrarily close by every point in its envelope". That still explains why its Minkowski dimension must be 2.
The article says that administrators are giving in to the demands of very involved, upper-middle class parents. What other incentives would an administrator have to keep low-income and minority students out of 8th grade algebra?
The Cyc project basically achieved what you're talking about, even without approaching AGI. They manually programmed concepts and relationships between things into a huge knowledge graph. Then they had heuristics for choosing the appropriate version of facts for a given context (e.g. level of rigor). It was arguably able to use a library of abstractions similarly to what Chollet is talking about, but couldn't learn new ones automatically through exploration or play.
It’s more horrible now. Syncing now opens a Finder window with an inconsistent look and feel, and sometimes fails to copy new songs in a synced playlist. The playlist view has the album art taking up half the screen, but there’s no way to shrink that section. And there’s no visual indication for whether shuffle is on - it has no grey box around it when enabled.
I kind of think they made it shitty on purpose to push everyone towards a subscription. Many of these issues apply to locally stored songs and playlists, which is how I use it.
Look at gravel or mountain biking! If you live near some farmland, ranches or a national forest, there are likely miles of public dirt roads that hardly get any car traffic. trailforks.com is a good place to start to see what's near you.
IMO this is the thing we should be scared of, rather than the paperclip-maximizer scenarios. If the human brain is a finitely complicated system, and we keep improving our approximation of it as a computer program, then at some point the programs must become capable of subjectively real suffering. Like the hosts from Westworld or the mecha from A.I. (the 2001 movie). And maybe (depending on philosophy, I guess) human suffering is _only_ real subjectively.
Imagine if avatar Gavin Belson had done the "metaverse legs" product reveal, with the animation running at 10 frames/second and a few legless avatars in the audience throwing up confetti. It would be almost too ridiculous to put in the show because the show itself would look like it was being cheap with the effects. But Meta had spent $36 billion on the metaverse at this point.
I don't think the self indulgence is unbearable until The Trial and "you stand accused of showing feelings of an almost human nature". The Trial is comedy/satire but since the rest of the album is serious and autobiographical, I roll my eyes at what a diva he sounds like there.
That's because Larrabee wasn't totally canceled, it was just pivoted from a gaming GPU to an HPC accelerator in the form of Knights Corner/Landing. The idea of having lots of x86 cores with wide SIMD units didn't change, but it was a lot more successful in the HPC world because anybody could compile an old MPI or OpenMP application with AVX512 and it just worked.
I do want to believe what you're saying because it means fewer life-years lost but... the other explanation is that covid plus the countermeasures to it really did impact people's health (not just the fatalities) and increased net mortality. We can check back in a few years but so far this explanation seems likelier to me.
When do you expect the death deficit to show up in mortality data? So far it looks like most (but not all) countries had an increase and then returned to the previous trendline: https://ourworldindata.org/grapher/number-of-deaths-per-year. For the world's total deaths per year, it definitely looks like the area under the curve increased due to covid.
If you like puzzles, look at The Talos Principle 2. It's mainly about asking whether humanity should even try to control the physical universe and spread life, or focus on sustainability and safety. The Culture itself referenced a couple of times in it.
L_R is not decidable. Deciding if x##y is in L requires simulating y until it produces |x| digits of output. x being anything and y = “while(true) {}” is a counterexample. Another way to look at it is that for any fixed x, "y produces x as its first |x| digits of output" is a nontrivial semantic property of y, so Rice's theorem shows that the property is undecidable.