It's a $20k, street-legal, EV modding platform. Sounds like you can mount your own infotainment system. Just an electric motor, battery, and chassis, and the rest is up to you. Isn't this what we've been asking for?
I play competitive Geoguessr at a fairly high level, and I wanted to test this out to see how it compares.
It's astonishingly good.
It will use information it knows about you to arrive at the answer - it gave me the exact trailhead of a photo I took locally, and when I asked it how, it mentioned that it knows I live nearby.
However, I've given it vacation photos from ages ago, and not only in tourist destinations either. It got them all as good or better than a pro human player would. Various European, Central American, and US locations.
The process for how it arrives at the conclusion is somewhat similar to humans. It looks at vegetation, terrain, architecture, road infrastructure, signage, and it just knows seemingly everything about all of them.
Humans can do this too, but it takes many thousands of games or serious study, and the results won't be as broad. I have a flashcard deck with hundreds of entries to help me remember road lines, power poles, bollards, architecture, license plates, etc. These models have more than an individual mind could conceivably memorize.
> Traditionally, coding involves three distinct “time buckets”:
> Why am I doing this? Understanding the business problem and value
> What do I need to do? Designing the solution conceptually
> How am I going to do it? Actually writing the code
> For decades, that last bucket consumed enormous amounts of our time. We’d spend hours, days or weeks writing, debugging, and refining. With Claude, that time cost has plummeted to nearly zero.
That last part is actually the easiest, and if you're spending inordinate amount of time there, that usually means the first two were not done well or you're not familiar with the tooling (language, library, IDE, test runner,...).
There's some drudgery involved in manual code editing (renaming variable, extracting functions,...) but those are already solved in many languages with IDEs and indexers that automate them. And so many editors have programmable snippets support. I can genuinely say in all of my programming projects, I spent more time understanding the problem than writing code. I even spent more time reading libraries code than writing my own.
The few roadblocks I have when writing code was solved by configuring my editor.
Difficult to describe them as choosing to do anything:
> ICE held the families incommunicado, refusing or failing to respond to multiple attempts by attorneys and family members to contact them. In one instance, a mother was granted less than one minute on the phone before the call was abruptly terminated when her spouse tried to provide legal counsel’s phone number.
What would they do, leave their child in an ICE facility and hope that somehow word gets back to family to go get them?
This is just a consequence of the fact that bfloat16 has a very high dynamic range which is not all used. People like hyperparameters that look like 0.01 not 10^10, even though there is the same fractional precision available at each exponent and if you multiplied everything - hyperparameters, initialized weights, training data, etc in a network by 10^6 things will still work more or less the same since the upper range is hardly used (with the possible exception of some small number of special functions).
Typical entropy of bfloat16 values seen in weights (and activations) are about 10-12 bits (only 65-75% or so of the value range is used in practice). Sign and mantissa bits tend to be incompressible noise.
This has been exploited several times before in the context of both classical HPC and AI, with lossless compression work from Martin Burtscher's lab (https://userweb.cs.txstate.edu/~burtscher/), fpzip from LLNL (https://computing.llnl.gov/projects/fpzip) and my library dietgpu from 2021 (https://github.com/facebookresearch/dietgpu) which we used to speed training on a large GPU cluster by about 10% wall clock time overall by losslessly compressing all data prior to send and decompressing upon receive (e.g., gradients, weights from backup, etc), which is still computing the same thing as it did before as it is lossless.
Also, rANS is more efficient and easier to implement in SIMD-like instruction sets than Huffman coding. It would reduce the performance latency/throughput penalties as well with DFloat11 (since we have to decompress before we do the arithmetic).
I am going to say a thing I say a lot: please edit Wikipedia. It is easier to do than you probably think! Wikipedia's biggest constraint is no longer money or server space, it's editor time (especially since LLM-based garbage is a force multiplier on disruptive editing that does not have a corresponding improvement to good-faith editing). Any topic area you know about and/or care about can benefit from your attention. Fixing typos is valuable. Adding photos is valuable. Flagging vandalism is valuable. Please edit Wikipedia.
For those unaware, the "SSL Added and removed here!" image is a reference to a diagram describing unencrypted communications between Google datacenters that leaked from the NSA in 2013 [1].
There is maybe nothing in the entire world that I am less sympathetic towards than the cause of font piracy / font liberation. You have perfectly good --- in fact, historically excellent --- fonts loaded by default for free on any computer you buy today. Arguing for the oppression of font licenses is, to me, like arguing about how much it costs to buy something at Hermès. Just don't shop at Hermès.
The latest dark pattern in all these apps is that they will mess up your order/ride and then "refund" you in the form of account credit. There's no way to actually get your money back. Trying to contact support results in an endless loop of help screens and chatbots. If you are able to figure out the magic combination of options that will get you to a support agent they will say "sure we will refund your money" but that is still only going to your app account. The last time I had an issue I literally had to ask them 3 times "I want you to confirm that you are refunding my credit card the amount it was charged" for them to finally agree to it. It is crazy what they can get away with.
This has moved heavily into group chats and I’m not sure it’s coming back.
Group chats are basically the Circles that Google+ saw the need for but could never get fully set up. A lot of people don’t want to share personal updates and photos to a broad swath of friends and acquaintances.
Meanwhile Instagram and Facebook keep evolving. Facebook is turning into a weird Reddit for older people. Instagram is turning into a hipper LinkedIn, where artists, musicians, and local businesses share career and business updates and advertise their wares.
The repository has been deleted. In addition, 26 other repos have been removed from the account. This is in line with DOGE members' quick response scrubbing data whenever put into spotlight, as previously seen with another "teen hacker". [0]
Good for him. This was an absolute ridiculous case. Lots of everyday items contain radioactive substances: old smoke detectors, uranium glass, old watches with radium dials, anti-static brushes, the list goes on and on. As a side note: coal power plants put quite a bit of radiation into the environment (technically 100x more than nuclear plants, if you sidestep the issue of waste), because coal contains Uranium and Thorium.
The amounts of Pu that were imported were not only minuscule, but also embedded in acrylic for display. As an alpha radiator, this is 100% safe to have and put on a shelf. You would have to completely dismantle it, crush the few μg of Pu into dust and then inhale it to be dangerous to your health.
I understand that people are afraid of radiation. I am too. However, it is important to know that radiation is everywhere all the time, and it is always about the dose. At the same time, we allow for instance cars to pollute the environment with toxic particulates that lead to many cancers, and somehow we accept this as unavoidable. But I digress...
For those interested, here's a video from "Explosions and Fire" on this issue, a channel I highly recommend anyway, this guy is hilarious: https://www.youtube.com/watch?v=M0JGsSxBd2I
Now that you got caught you are fixing it and writing fancy PR fluff. An org the size of MS should have clear policies and processes of how to handle open source forks like this. Unless we assume “bad faith” here. This is a pretty bad look.
I wonder how many other projects are not attributed correctly. Are you checking up on them also or just waiting for the next HN post?
That said, the author of Spegel should have used another license if he wanted more “recognition” or the like.
Pope Francis caused quite a bit of controversy among Catholics. From his crackdown on the TLM (Traditional Latin Mass) to his often unscripted, pastoral tone on issues like sexuality, economics, and interfaith dialogue, he unsettled many and yet drew others closer to the Church. With his passing, we’re left to process a papacy that disrupted in the deepest sense of the word.
As a Catholic, I often found myself both inspired and unsettled by him. His theology wasn’t always systematic, but it was deeply Ignatian, rooted in discernment, encounter, and movement toward the margins. Francis often chose gestures over definitions, and presence over proclamations. That doesn't always scale well in a Church that spans continents, cultures, and centuries.
His legacy will be debated. But I think what made him so compelling, especially to someone who lives in the modern world but tries to be formed by ancient faith is that he forced us to confront the tension between tradition and aggiornamento not as an abstract debate, but as something lived.
He reminded me that the Church isn’t a museum, nor is it a startup. It’s something stranger.. the best I can described it is a body that somehow survives by dying daily.
- Requiem aeternam dona ei, Domine, et lux perpetua luceat ei. Requiescat in pace. Amen.
It should be forbidden for all device manufacturers to make apps, tracking, registering, etc. mandatory.
Every TV, phone, camera, tablet, fridge, ... is becoming a spying device like in the worst scifi dystopias. And as soon as the company stops supporting them they become trash to pollute the planet so they can sell you the next one.
Regulations should have come a decade a go. We own nothing, we have no privacy, we are sold products 24/7. I will vote for a goverment that protects me of this total corporate surveillance. It is their duty towards citizens to do so.
And it will happen, like feudalism died this techno-feudalism will die too.
We can finally just take a photo of a textbook problem that has no answer reference and no discussion about it and prompt an LLM to help us understand what's missing in our understanding of the problem, if our solution is plausible and how we could verify it.
LLM changed nothing though. It's just boosting people's intention. If your intention is to learn, you are in luck! It's never been easier to teach yourself some skill for free. But if you just want to be a poser and fake it until you make it, you are gonna be brainrot waaaay faster than usual.
Cybersecurity is not my main field but this sounds beyond suspicious.
> Berulis [...] and his colleagues grew even more alarmed when they noticed nearly two dozen login attempts from a Russian Internet address (83.149.30,186) that presented valid login credentials for a DOGE employee account — one that had been created just minutes earlier. Berulis said those attempts were all blocked thanks to rules in place that prohibit logins from non-U.S. locations.
> “Whoever was attempting to log in was using one of the newly created accounts that were used in the other DOGE related activities and it appeared they had the correct username and password due to the authentication flow only stopping them due to our no-out-of-country logins policy activating,” Berulis wrote. “There were more than 20 such attempts, and what is particularly concerning is that many of these login attempts occurred within 15 minutes of the accounts being created by DOGE engineers.”
Somehow each paragraph reveals something even worse than the last.
> Berulis [...] and the associate CIO were informed that “instructions had come down to drop the US-CERT reporting and investigation and we were directed not to move forward or create an official report.” Berulis said it was at this point he decided to go public with his findings.
> {0} initializer in C or C++ for unions no longer guarantees clearing of the whole union (except for static storage duration initialization), it just initializes the first union member to zero. If initialization of the whole union including padding bits is desirable, use {} (valid in C23 or C++) or use -fzero-init-padding-bits=unions option to restore old GCC behavior.
This is going to silently break so much existing code, especially union based type punning in C code. {0} used to guarantee full zeroing and {} did not, and step by step we've flipped the situation to the reverse. The only sensible thing, in terms of not breaking old code, would be to have both {0} and {} zero initialize the whole union.
I'm sure this change was discussed in depth on the mailing list, but it's absolutely mind boggling to me
Not sure if it's being exaggerated for comedic purposes but it is interesting to me how alien the act of sending a letter by post is to the author. Granted I don't send them very often but I wouldn't think much of it if I had to. But I guess younger people and particularly those in tech may genuinely never need a reason to send a letter (or, it seems, write an address by hand).
A key point here, which the judge brought up with the ICE agents, is that they only had an "administrative warrant".[1] An “ICE warrant” is not a real warrant. It is not reviewed by a judge or any neutral party to determine if it is based on probable cause.
"An immigration officer from ICE or CBP may not enter any nonpublic areas—or areas that are not freely accessible to the public and hence carry a higher expectation of privacy—without a valid judicial warrant or consent to enter."[2]
The big distinction is that an administrative warrant does not authorize a search.
Hi Philip, I'm Lachlan from the Cloud Native Ecosystem team at Microsoft. Our team works in the cloud native open-source community with a goal of being great open-source collaborators in these projects and communities, and I’m sorry that this happened.
We appreciate your leadership and collaboration on Spegel and see your project solving a real challenge for the cloud native community. I wanted to thank you for your blog post https://philiplaine.com/posts/getting-forked-by-microsoft/, let you know what we’re doing, and address a few points.
We’ve just raised a pull request https://github.com/Azure/peerd/pull/110 amending the license headers in the source files. We absolutely should have done better here: our company policy is to maintain copyright headers in files – we have added headers to the files to attribute your work.
I also wanted to share why we felt making a new project was the appropriate path: the primary reason peerd was created was to add artifact streaming support. When you spoke with our engineers about implementing artifact streaming you said it was probably out of scope for Spegel at that time, which made sense. We made sure to acknowledge the work in Spegel and that it was used as a source of inspiration for peerd which you noted in your blog but we failed to give you the attribution you, that was a mistake and I’m sorry. We hear you loud and clear and are going to make sure we improve our processes to help us be better stewards in the open-source community.
Thanks again for bringing this to our attention. We will improve the way we work and collaborate in open source and are always open to feedback.
> He accused Dugan of “intentionally misdirecting” federal agents who arrived at the courthouse to detain an immigrant who was set to appear before her in an unrelated proceeding.
It sounds like the arrest isn't because of any official act of the judge, but rather over them either not telling the ICE agents where the person was or giving them the wrong information about their location.
There are some pretty broad laws about "you can't lie to the feds", but I think the unusual thing here is that they're using them against a reasonably politically-connected person who's not their main target. (They're normally akin to the "we got Al Capone for tax evasion" situation -- someone they were going after, where they couldn't prove the main crime, but they could prove that they lied about other details.)
EDIT: since I wrote that 15 minutes ago, the article has been updated with more details about what the judge did:
> ICE agents arrived in the judge’s courtroom last Friday during a pre-trial hearing for Eduardo Flores Ruiz, a 30-year-old Mexican national who is facing misdemeanor battery charges in Wisconsin.
> Dugan asked the agents to leave and speak to the circuit court’s chief judge, the Journal Sentinel reported. By the time they returned, Flores Ruiz had left.
I.e. the ICE agents showed up in the middle of a court proceeding, and the judge said they'd need to get permission from the chief judge before they could interrupt proceedings. The judge then didn't stop the defendant from leaving once the proceeding was done.
EDIT 2: the Milwaukee Journal-Sentinel article says:
> Sources say Dugan didn't hide the defendant and his attorney in a jury deliberation room, as other media have said. Rather, sources said, when ICE officials left to talk with the chief judge on the same floor, Dugan took the pair to a side door in the courtroom, directed them down a private hallway and into the public area on the 6th floor.
Which is an escalation above the former "didn't stop them", admittedly, but I'm not sure how it gets to "misdirection".
That is an absolutely brilliant turn of events – strong evidence that the font in an anti-piracy campaign was itself arguably a copyright-infringing knock-off.
Someone should sue FACT for copyright infringement – and refuse to settle.
1. It isnt clear ICE agents have any legal authority to demand a judge tell them anything. 2. It is highly likely this is an official act, since it would be taken on behalf of court, so the immigrant can give, eg. testimony in a case.
A "private act" here would be the judge lying in order to prevent their deportation because they as a private person wanted to do so. It seems highly unlikely that this is the case.
I find this type of problem is what current AI is best at: where the actual logic isn't very hard, but it requires pulling together and assimilating a huge amount of fuzzy, known information from various sources
The US is falling way behind in electric vehicles. If BYD could sell in the US, the US auto industry would be crushed.[1]
What went wrong is that 1) Tesla never made a low-end vehicle, despite announcements, and 2) all the other US manufacturers treated electric as a premium product, resulting in the overpowered electric Hummer 2 and F-150 pickups with high price tags. The only US electric vehicle with comparable prices in electric and gasoline versions is the Ford Transit.
BYD says that their strategy for now is to dominate in every country that does not have its own auto industry. Worry about the left-behind countries later.
BYD did it by 1) getting lithium-iron batteries to be cheaper, safer, and faster-charging, although heavier than lithium-ion, 2) integrating rear wheels, differential, axle, and motor into an "e-axle" unit that's the entire mechanical part of the power train, and 3) building really big auto plants in China.
Next step is to get solid state batteries into volume production, and build a new factory bigger than San Francisco.
The people configuring WAF rules at CDNs tend to do a poor job understanding sites and services that discuss technical content. It's not just Cloudflare, Akamai has the same problem.
If your site discusses databases then turning on the default SQL injection attack prevention rules will break your site. And there is another ruleset for file inclusion where things like /etc/hosts and /etc/passwd get blocked.
I disagree with other posts here, it is partially a balance between security and usability. You never know what service was implemented with possible security exploits and being able to throw every WAF rule on top of your service does keep it more secure. Its just that those same rulesets are super annoying when you have a securely implemented service which needs to discuss technical concepts.
Fine tuning the rules is time consuming. You often have to just completely turn off the ruleset because when you try to keep the ruleset on and allow the use-case there are a ton of changes you need to get implemented (if its even possible). Page won't load because /etc/hosts was in a query param? Okay, now that you've fixed that, all the XHR included resources won't load because /etc/hosts is included in the referrer. Now that that's fixed things still won't work because some random JS analytics lib put the URL visited in a cookie, etc, etc... There is a temptation to just turn the rules off.
