It is such a big category error to say "we're building simple apps without looking at code, therefore no one reads code anymore."
I'm not sure if this comment is meant as a little joke or observation, but I keep on seeing serious comments like this. It is serious anti-intellectualism to scoff at people reading programming books or understanding things.
Some other great low flying titles:
The Computational Beauty of Nature by Gary William Flake
https://mitpress.mit.edu/9780262561273/the-computational-bea...
The Nature of Computation by Cristopher Moore and Stephan Mertens.
https://nature-of-computation.org/
Algorithms by S. Dasgupta, C.H. Papadimitriou, and U.V. Vazirani
https://cseweb.ucsd.edu/~dasgupta/book/index.html
and On Lisp, of course:
Of course, lots of things that make modern software bigger are things we very much don't want to do without--full Unicode support, accessibility, clever rendering, improvements to quality of life. But our 'foundations' are also built higher; the Web platform and a bunch of native toolkits provide a lot out of the box. (The platform doesn't include all the widgets that are now quasi-standard, of course, so we build a lot on top anyway.)
There are also a ton of useful third-party libraries you can call on without getting too bloated, and there's a lot that would have been bloated then but not now: in the 90s you probably wouldn't embed an SQL DB you didn't really need, but SQLite is a pretty safe dependency now.
Nothing in particular I'm really driving towards here but I think "how do I keep the code to functionality ratio good" can still be an interesting angle from which to look at high-level design decisions even as a lot of the circumstances and numbers have changed over decades.
I'm building Retro Song Bird, a speaker that plays the soundtracks from your classic Game Boy and Game Boy Color games. There were a ton of unique technical challenges to building the prototype in a way that could play instantly, have cool features, and not attract lawsuits from IP holders! I've discovered a ton about the artistry of these early game music composers and the technical limitations imposed on them. I've also discovered a ton of hidden bangers as I collected and tested 150+ carts. I'm most excited to see how people make music using the stems of classic tracks or perform DJ sets with two beat-synced carts. Happy to answer any questions, thanks!
I disagree. The article is very clear about the distinction, especially during the segue into the more substantive issues with the writing. And it does put due blame for the typos on the editors, not just the author.
Perhaps the most important clue is in the first paragraph: “one thing I can always rely on is being let down by a popular or acclaimed piece of contemporary art.”
It’s hard to be a contrarian unless you’re willing to go to some lengths to find something to be outraged about.
These days lots of books are self published. And even the big publishers spend less money + time on this kind of thing.
I hear that excuse constantly on the web if I try to correct people. And maybe it doesn't "prohibit access", but it does force me to think about things that aren't the point of the message, in order to figure out the point of the message.
I've given up correcting people. I'm old enough now that it's time to just watch people's choices come back on them. The whole world will pay for it, and I've done what I can.
I can't even get people to stop using "weary" to mean "wary". That's a whole different meaning, and there are so many people that use it wrong.
Way ahead of you boss!
"But horizontal wealth means not letting your increased income dictate your tastes. You like books and now you have money? Buy more books! Change those catenary bookshelves for good hardwood ones! In my case, build a library extension to your office. And, of course, you buy what will be useful for that most wonderful of pursuits, blind research, which is research without direction for the sheer joy of it."
(Original source is his book A Slip of the Keyboard - this linked blog post quotes that bit.)
I wonder how my life would have turned out differently. I am a nerd. I am an introvert. I am single. Mostly, I am enjoying it. Sometimes - and increasingly more often - I am questioning it. I have baggage. Was visiting my (younger) sister's grave today. She had baggage with a sad outcome. I am thinking more about my existence.
I wish this world would be a more kind place. And I am praising anyone who is contributing to that. Try to be a contributor.
One distinct memory stands out. I could not wrap my mind around the Object-Oriented extensions on Pascal. I read, re-read, and re-re-read the section of the book describing what polymorhpism is and how it worked, probably at least a dozen times. Somehow it seemed really important and I kept coming back to it. I worked the examples. And then one day the ah-hah moment arrived. I remember it felt like a shock, why it was useful and what I could do with it. I promptly used my knowledge to write a new game using OOP principles to guide the rudimentary "characters" of various geometric shapes at the core of the game.
Fun times, and very formative to me!
And: "A Personal History of Compilation Speed, Part 2" https://prog21.dadgum.com/47.html
He offered three courses. CompSci 1 used Turbo Pascal and Delphi. CompSci 2 used C++ and Java. CompSci 3 used Microsoft Turbo Assembly and a language of the student’s choice.
We had to write our code and save them to floppy discs for submission since the lab systems had DeepFreeze installed on Win95, so any local storage wouldn’t persist.
Perhaps "The End of Microsoft Research's AI Frontiers Group" would be a less ambiguous title.
Signal is incredible user friendly, privacy respecting, the gold standard for encryption, cross platform, etc. Yes, it’s centralized, yes it’s imperfect in many ways, but it’s easy and it’s pretty damn good.
If you want to help the people around you to become more involved in understanding and owning their tech, it has to be an easy and good experience. It has to be tangibly better, not just abstractly better with tangible downsides.
I use XMPP, but would never try to convince my friends and family to use it; it’s too complicated, the user facing apps all suck (Snikket and Cheogram are the best of the bad lot), and the incentives to use it aren't there. As you said, it’s just not attractive to a regular user, since it’s all around a worse experience than what they currently use. The one exception is that I do occasionally recommend XMPP for SIP access via jmp.chat to some of my more technical friends who have a tolerance for clunky user experiences in favor of privacy and affordability. I am holding onto an unreasonable hope that Signal or Whatsapp or others will consider XMPP+OMEO in the future to popularize the standard, but that is extremely unlikely.
Matrix also is an unfortunately terrible alternative to discord (even Element is a rather lacking user experience), so while I have pushed to move groups there before, I no longer recommend Matrix. I suggest signal or nextcloud as alternatives when possible; the remaining large groups who need the closed server forum style system are usually unlikely to move to Matrix anyway.
Unless everyone in your circle is on HN all the time, you will probably fail to convince them to change.
The more likely outcome is that you wind up annoying everyone involved and don't get invited to the new discord server.
- Whether family or friends, always suggest it as a "fallback" for when "preferred service" is down and to use when discussion ultra-sensitive topics or sharing financial files. With time they may just stay connected to it if the friction to do so is low. It has to be their choice, not yours.
- What helps? Add more incentives such as perks that "preferred service" does not offer. No file size limits. No or few file type limits. No content filters. Add emoji's or gifs that align with their inside jokes and their style of humor. Keeping random people away is very important. Making your friends and family administrators or moderators and let them decide how it operates. Do not moderate or power-trip over your friends or family or they will leave. Limit rules to "Do not break the laws of this country in public chatrooms" and private chat rooms if you do not physically host the server. Database and file backups become even more important but they will feel in control of a thing. Have a keyword or code word they all learn means to go check the private platform that "primary platform" would not recognize. Maybe when they see "Turtle Farts!" they know to go log into your service.
- They may just stay on their primary preferred service. Every group is different and not every group will see the benefits of your fallback service. They may not appreciate the work you put into setting it up for them. Do not expect to get the friends of your friends and family to move to your service. They may not appreciate the OpSec value until "Bad Things"(c)™2038 happen to enough people. Just be there for when it does.
- If there are concerns that you would spy on them, consider instead using eJapperd or Prosody as you can implement OMEMO for individual and group end-to-end encrypted messages. Set up TLS fingerprint checks from multiple locations to reduce the risk of MitM attacks until XEP-0474: SASL SCRAM Downgrade Protection is fully supported and active by default on both servers and all clients. This is the only thing holding me back from setting up a public Prosody server as 0474 is still experimental. See [1]
- Avoid (block) federation on private servers for friends and family otherwise it is not a private server.
If I were you, I would probably ensure that the fridges you used have residential or commercial (aka non-DC) IP addresses. Maybe Shodan uses this data for their "honeypot" label, my experience was some time ago so no idea though.
The directors always try to avoid this criticism by saying, "Well he goes to jail in the end," but I think it's a bit of a cop out.
I tend to agree with the idea that art should try to embody some moral values, which goes against a lot of contemporary thoughts on aesthetics. But I think it’s a basic fact that people look toward fictional characters as role models, directly or indirectly. Avoiding that responsibility is not good.
I did not know this! Thank you for posting!
These days we'd be lucky to see certain models discounted to $314.
Its an interesting comparison that Im not sure reflects positively on the RISC-V C906 core when you factor in core area and process nodes and core area, 22nm vs 250nm. Even if we choose the worst case scenario for the P3 cores to compare which would be the Katmai core. We are basically look at ~parity with a 30 year old core design
C906 core only @ 22nm = ~0.6 mm2
C906 core only @ 250nm = ~78 mm2 (geometric transform)
P3 core only @ 22nm = ~0.5 mm2 (geometric transform)
P3 core only @ 250nm = ~76 mm2
Apologies if this is messy and badly formatted. Im currently at a wedding, sat waiting between the ceremony and the after party and Ive quite a few beers. I responded because its something I looked at previously with RISC-V cores and I already had the answers in my head
In the 2020s or somewhere before that we got a bunch of dedicated cheap MCUs that were good enough and I think stole the thunder from full Linux but there are still projects to run it on things like the ESP32 family.
Either way, hardware hacking is still my favorite hobby to watch.
BUT! I feel that lost hope could/will turn into a new forms of "min-maxing" or whatever where we will see people focusing on small, efficient software to make the most of what we have to weather the drought. We got spoiled with cheap hardware. Time to make the most of all that cheap hardware we have.
Though of course, replacing this hardware as it breaks down will be expensive. Perhaps we could go back to 1/2/4 GB RAM sticks and 256/512 GB Disks. Then run minimal Linux installs, minimal GUIs, minimal services, minimal software.
To amend this: Reading the 8088 domination article and the little $8 RISC-V Linux board I think we have plenty of opportunity. Maybe we learn to shed our inefficient BIG computers for small, low power, purposeful, yet general computers. I play with Plan 9 a lot and its portability, small footprint, low resource, and distributed design has showed me we can do a lot with a little as long as we stay disciplined to keep things simple and to the point.
Not everyone subscribes to the "cloud compute is more important than everything else" outlook, but everyone is affected by it nonetheless.
Forget about AI, the entire PC market is getting killed over this.
As an avid gamer I'm unsure if in 5 years I can even buy a gaming PC anymore. Not even because of cost, but it's unclear if anyone will still make the components anymore.
NVIDIA isn't interested, nor is AMD.
Apparently the last three times this cycle happened, the manufacturers that built out capacity went bankrupt when it popped, and the ones that didn't didn't, so they're actually being wise.
Will we go back to PCs? As in: are we going to get cheap local hackable frontier AIs at some point? I hope so.
Good god the internet is so sensitive nowadays
Projects like these are defiling the programmer's guild.
This is a better research direction I feel
https://research.google/blog/the-sleep-framework-a-new-way-f...
So the main improvement to memories I'd like to see is better judgment on which ones to create.
The parts about "dreaming" make me think of how agent transcripts contain a ton of information, which I think is currently under-utilized. I'll repeat advice I've seen here a few times: ingest your own transcripts into duckdb (claude code will build that pipeline in about 10 minutes) and do some data analysis on them. Some ideas if you don't know where to start: look at tool call failures to see if you can prevent common ones, check out if you have large cache misses and what causes them, look at various efficiency metrics over time (tokens per $, lines of code per tokens, etc.) with main thread vs sub-agents, etc. I started with this sort of things, and ended up following some threads to interesting insights. For example, I found indications that I tended to make Opus 5.5 / Sol 6.1 divide their work into units that were too small. They have the capacity to work on more stuff at once than older models, and I wasn't really taking advantage of that.
I lost all my respect for isaacson at that time as it was clear he’d entered a bubble himself.
This also seems along with a lot of other recent moves to reveal that money is getting tighter over there.
The U.S. is being protectionist because of Chinese cars that can spy on U.S. soil-- just do the right thing and stop all the car spying nonsense. Require manufacturers to open the code and let us even develop alternatives.
Tell GM.
While most others are wasting money on failed forays into the EV market, Toyota has quietly made their entire fleet (modulo trucks) hybrid.
At this point anyone can design a chassis to accommodate a skateboard battery and some drive motors. The only thing Toyota would need to wholeheartedly jump into EV is the supply chain and production capability for big battery packs. Admittedly this is a biggie, but don't discount the possibility that batteries are not at the end of their evolution and early entrants might get burned by some new chemistry or form factor.
The model Y is still the best selling EV despite the wide choice of other brands. BYD out sell them in aggregate, but Tesla are holding their own pretty well. Of course, all the Tesla cars here are made in the Shanghai Gigafactory, but still, I don't think Chinese brand dominance is a forgone conclusion. I think the real danger is if American protectionism makes Tesla feel like they don't need to innovate any more.
The thing is that ICE vehicles are so inefficient to begin with, but a towed trailer needs the same amount of energy regardless of what is towing it. Imagine a trailer takes 1 unit of energy to move, and ICE truck takes 5 units to move itself, and EV truck takes 1 unit to move itself.
The ICE truck towing it would go from 5 units to 6 units, reducing the range by 16%.
The EV truck would now go from 1 unit to 2 units, reducing the range by 50%.
So the better your efficiency, the worse you are at towing. Ideally EV trucks would have such a huge battery that even when the range is reduced by 50%, it is still as long as a gas truck's range, but modern battery technology is still too sucky to achieve this.
I'd say Koreans are probably the closest to the Chinese, but Japanese manufacturers have fallen behind. Toyota spent too much time with hydrogen. Nissan? Honda? Mazda? None of them have a solid lineup of electric vehicles.
I appreciate Toyota's hybrids, but that is not the future and every model year that they try to delay, they are missing out on developing the technology, experience, internal expertise, and platform refinement.
The Cybertruck I thought was initially kind of neat but when you try to use one as a truck you quickly can see some glaring problems that most people who buy a truck want. Towing with an EV is terrible. I've heard there's power stations with pull throughs but havent' seen them personally. The truck bed isn't practical for anything but a lawn mower. When I ride in one it rides like a car. It doesn't look good dirty. I just think they were going for warthog from Halo and thought the rest would work out. It's also super expensive for all these issues. It's a hobby vehicle. A toy.
I think Tesla is struggling personally. The marketing tactics of showcasing cars we'll never be able to buy just to cause a scene. Don't get me wrong the new Roadster my jaw dropped and I agree with Elon now that SpaceX and Tesla are the 'same' company I think it's great but let's be realistic. Most people want cars that are cheap, dependable and sort of cool. Not everyone can afford multiple cars, and a corvette that isn't feasible. Like I said I work for a competitor (GM) and I fear China is going to eat our lunch. Unless we make infrastructure for charging better and make it a no brainer.
30kW is a rate.
30kWh is a measure of energy, about 100MJ.
30kWh/day is a rate again, 100MJ/day.
It seems clunky, but it's a way energy usage in a house is typically measured.
My Ioniq 5 has been a godsend. No, it can't power the house without dorking around with my electrical panel. But we can run the fridge, dishwasher, a few standing lights and keep phones and computers full charged.
I used a similar setup to Alex's but with the outlet in the rear seat instead of the V2X connector. I will probably by the V2X connector since it supposedly hanldles higher power draw.
Tesla forecasted selling 250,000 trucks per year by 2025
Actual sales
2024 - 38,965
2025 - 20,237
2026 to date - 11,586
This includes trucks that have been “sold” to other Musk companies.
This product was a complete failure, making the Ford Edsel look like a roaring success
My experience with power outages (after big summer storms, not a hurricane thank goodness) has been that my neighborhood is impacted, but that places nearby have power. They tend to restore commercial first, so the supercharger near me is up and running before my house.
- https://www.theautopian.com/the-average-full-size-truck-now-...
- https://www.ceicdata.com/en/united-states/new-vehicle-averag...
The cybertruck was built in an era of high truck prices driven by zero interest rates. It bet on the price of trucks going up, just as the Ford F150 electric did. That was the wrong bet.
If it was a $40,000 truck, it would have a much bigger market despite all the other explanations of its failure. The market moved under it.
But it is still cheaper to stockpile some generators and gas than stockpile a cybertruck. If there's an EV charging station up within a 1 hr drive, gas stations are also going to be up.
> Most of the disasters I'm aware of people either get their power back in a short enough amount of time that they can use a normal generator for the gap
TBF an EV car has an absolutely massive battery relative to the electric consumption of a house, unless you're using resistive heating all around. Especially if you ration use.
This is a pretty common feature in High capacity cars.
Even my mid-range BYD came with power strip attachment so you run basic electrical appliance with it.
It's a luxury vehicle that beats SUVs hands down, and most functions of urban trucks as well.
It just gets soccer mom vibes, so it's not $120,000
I am not a fan of the truck, but let’s make some sense here
Of course the Prius Prime on electricity is way more efficient than both, but it still uses a lot more energy than a house.
I got one, years ago. The bed was not large enough to readily hold full sheets of plywood (the minivan I owned before could hold full sheets!). And yet, the truck was long such that parking, maneuvering in parking lots, etc, was a pain in the ass (here the minivan was about tied with the truck).
So sucked as a regular errand-car, sucked for hauling sheet goods.
(Sold it during the tail of Covid when used car prices were skyrocketing for whatever reason. Never again.)
I think this is why these EV trucks don’t do well (ford also cancelled the F150 lightning). For truck things, they aren’t particularly capable (you’d get less than 100 miles towing). For daily driving duty, a car or small suv is much more efficient. You can power your home during a storm from most EVs these days, you’d just have to top it up more often
They want ~9k for a freaking 13.5kwh pack.
I can buy a UL certified, 16kwh, outdoor rated pack for 3.5k (ex - EG4 allweather)
I can buy just the LFP batteries themselves for the new powerwall 3 LFP version (no BMS, no shell, no certification) for $1116... https://18650battery.com/products/byd-302ah-3-2v-lifepo4-pri...
Powerwall is a scammy product. WAY too expensive for the storage it provides.
---
So the summary is: Cybertruck is using a different battery chemistry, Powerwall is a scam.
(I mean - to be clear, the Cybertruck is also a scam, but it's a different kind of scam not directly related to the battery capacity, since it's hauling around a 123kWh NCMA battery pack)
However, it provided content and distraction and kept the stock high longer than it should have and I feel like that’s what Musk continues to excel at.
An $80K-100K car that replaces the two isn't some genius purchase.
Sure you can spend all the money you want on a luxury toy, nothing wrong with it, but it's weird trying to justify it as anything but that.
I don't think the residential power generator is needed anymore if you have an electric car.
https://www.tesla.com/powershare/#vehicles
> Powershare allows you to draw power directly from your compatible Tesla vehicle to power your devices, your home or the grid.
> Powershare is currently available for all Cybertruck vehicles, and all new Model 3 and Model Y2 vehicles ordered on October 1, 2026 or later.
> To verify if your vehicle has Powershare, navigate to ‘Software’ > ‘Additional Vehicle Info’ > ‘Powershare Support: Enabled’ on your vehicle touchscreen.
48 volts, steering wheel, no paint ( better durability), trunk, wipers etc. the car pushed the boundaries. If you compare level of creative thinking of cybertruck vs lets say electric ford, it's not even comparable. I respect that despite one person behind it. I don't care how it looks. It's polarizing. oh well.
My 11 y.o. relative from abroad asked me to take rl pictures of the truck, after he heard my neighbor bought it.
It probably would have been better had they not been mass produced, and maybe did a very limited quantity run - make it more like a collector item... I suppose thats what it'll be now. I see quite a few on the road, it takes away from the uniqueness.
It’s unclear to me how a vehicle battery can be 9-10x the size of a house battery, be used as a house battery, include an highway safe cabin, motors and wheels and only cost 4x what a single home battery costs. Oh, and at least Kia rates their battery warranty based on miles driven, so using it as a home battery does not impact any 10-year capacity guarantees.
AI tells me they are completely different though, so I must be missing something.
Mostly I've seen younger kids love it. I was at the park and a little girl saw a Cybertruck pull up -- "Look, Dad, it's from the future!"
I think Cybertruck is probably ultimately a failure, but the amount of hate it gets when they are at least trying to push things outside of the normal makes me sad. The amount of negativity towards anything new that's not perfect is insane.
The battery capacity on those things is equivalent to 9 Tesla Powerwall units or something close to it.
They may be ugly, but the emergency prepper in me always wanted one after hearing that story. Before, I always wanted to install powerwalls...but I never considered in an extended event being able to drive my powerwalls somewhere to recharge them.
GM - $74.5B market cap (6.2M cars sold annually)
Ford - $50.2B (4.6M)
Volkswagen - $40B (9M)
BYD - $100B (4.6M)
Toyota - $256B (11.3M)
Hyundai - $57B (7.3M)
Tesla - $1.4T (1.6M)
The numbers simply don't add up.
They are going to be very valued collectors items in 30 years once the kids of today have enough disposable income to buy whichever are left roadworthy by then.
Try two generations. I think it was obvious since the oil shocks that fossil fuels were not a good bedrock of the economy.
It's not every quarter, but it's steady and that spells absolute disaster for US car manufacturers:
https://www.coxautoinc.com/insights/q3-2026-ev-sales-report-...
US car companies have been protected and bailed out too many times and are now not only completely uncompetitive on the global market, but set up to be completely demolished as EVs take over.
Their bad leadership and overall inefficiency mean that either investors are going to pay a massive price, or US consumers are going to pay a massive price. If we keep on pushing up consumer costs by propping up bad manufacturers with protectionism and bail outs, the US's entire economy is drug down by that.
There needs to be some creative destruction, ruthless culling of the least efficient manufacturers, so that their best workers can go to the better manufacturers, if the US is ever going to be a top tier car manufacturer again.
I hadn't followed just how bad US manufacturers were failing at EVs the past two years, what with all the other disasters unfolding, thanks for the wake up call. Two years of wasted time, of the US falling even farther behind... such a massive fail and yet not the most massive.
hilarious sentiment
Not so in Europe, where the top 10 EVs sold in 2025 [1] has not a single Chinese car.
[1] https://autorai.nl/tesla-van-de-troon-gestoten-als-populairs...
That EV market share direction in the US, though .. really bad compared to non-petrostates (except Norway)
- The emissions credit market (where Tesla made most of its money) is dead due to lowered standards.
- Manufacturing costs haven't come down.
- EV sales share in the US is down to 6% (and falling every quarter).
- Chinese EVs are dominating every market outside the US.
- All the overhyped moonshots (sub-$30K car, unsupervised autopilot, autonomous robots) are nowhere to be seen.
It's hard to see where Tesla goes from here as a company. The stock is still surviving on Elon hype but has been stagnant for 5 years. The day he gets bored and decides to resign to focus on SpaceX full time the company is basically done.
A bad design that costs too much, the only people who bought these were ideologues committed to the vision of a man responsible for millions of unnecessary deaths, the gutting of the American state, and some of the most destructive politics of hate in living memory.
I celebrate their passing.
I still remember years ago seeing one in a grocery store parking lot, with a banana yellow wrap and the Zelda unlock sound as its lock chime. A brief moment of inconsequential joy.
Lastly, having the cyber truck drag race a Porsche 911, while towing a Porsche 911 (and win) has to be one of the best marketing stunts of all time.
Musk has enthusiasm, yes, but he’s let his luck at PayPal and successfully buying Tesla from the original founders go to his head. Nobody is right all of the time, and no serious project goes perfectly according to the original vision without adjustments. There’s nothing wrong with the idea of an electric truck — companies with better engineering culture are making great ones — but the Cybertruck failed exactly the way everyone who isn’t a superfan expected: it’s designed in a way which guarantees high costs and it just doesn’t deliver corresponding benefits for most people. Diving into the culture wars against a large chunk of potential Tesla buyers was a bad idea but even if he hadn’t, there just aren’t that many people who have that kind of money to spend and the ones who do usually won’t accept so many limitations and flaws.
They could’ve done something like the Slate truck a decade earlier.
The end result is a vehicle with all of the costs of a normal unibody design plus the costs of dealing with cutting, folding, and attaching thick sheets of stainless steel.
It never delivered the promised $40k starting price or the promised 500 mile top spec range either.
Also painting IN THE FACTORY isn't expensive. Re-painting the part on repair is because of all of the prep work required (take apart, clean, color match etc).
Might be break-even, but I doubt it would be cheaper
"It will be cheaper" is a perspective of ignorant "smart guy" that thinks because he got successful in one thing that they can weight in on industry processes they have no idea about
ToothPaste allows a user to transmit AES-256 encrypted keyboard and mouse commands to any USB-compatible device wirelessly, without the need for specialized drivers or extensive set-up using WEB-BLE, a Cryptographic IC and an ESP32-S3.
Though to be fair the concept of a powerful city lost to time is something that resonates a lot, and there's multiple similar myth like the isle of Ys, the lost continent of Lemuria, or in the arabic folklore 'Iram of the pillar', often called 'Atlantis of the sand'.
[0] There was an episode on it in the excellent podcast 'In Our Time' by the BBC (https://www.bbc.co.uk/programmes/m001c6t3)
I found Fira (nearby "large" town on Santorini, very close to the sites) miserably crowded, but seeing Thera was very cool. This is all about 1500 years older than Pompeii, so a very different experience IMO.
People, please try to understand what you are building. Don't just build it as asked!
Despite the title, the article is not making claims of, or demonstrating general usage improvements or general coding agent improvements
Update: let me elaborate more, I read the HN entry about the team of 5 Qwens being close to fable quality... so I embarked to add to my harness (https://github.com/alainnothere/privibe because we all have one right?) and once I had added the orchestrate tool I had to now measure somehow it it's able to improve quality, so I claude and then Qwen to run a experiment to call my harness passing the model to use and ask them to do something, 3 rounds I think and see what quality came out of it... and as per the grading... qwen and claude agreed that the quality of the response comming out the harness using the orchestrate was VERY CLOSE, with opus and fable providing more "in my opinion you should" which matches my experience on why they are better.
One more update: I can't get copilot models to work decently, it may be my way of working... but I can get Qwen to follow what I try to do... if Qwen get's me he can then fire copilot cli and get amazing results by using copilot... bonus points seems like copilot cli is cheaper with the AIC (cost of the requests) so you can make the copilot subscription longer
It's like a better alternative to xterm.js, and better than existing terminal emulator libraries.
If you have structs that might need to be extended, you can use "versioned structs". One popular technique is to include the struct size as the first field. The struct size is set to `offsetof(mystruct, lastmember) + sizeof(lastmember)` and the receiver of the struct can use this to check if a member exists. The ARA API is a good example: https://github.com/Celemony/ARA_API/blob/65ec5c43b943a48cb54...
What really caught my attention was this: "So from the start, I took inspiration from C libraries I enjoyed working with and designed the API to have a strong, stable ABI that allows for future additions." What are good examples of this? I feel like I just don't have enough experience with well-designed C libraries to know what to look for.
Mitchell could have easily snuck his competing product Rex in there but resisted. Respect.
Can anyone give a comparison to libvte? There are many terminals that are just thin wrapper around that (gnome-terminal, xfce4-terminal, mate-terminal, lxterminal, sakura, …). It is tied to GTK, so it has a much narrower scope. But still it feels related in spirit.
People linked to them came up with the "prediction" nonsense to sound modern and technological as a way to dissociate themselves from the life/family destroying gambling of old.
But they are gambling, pure and simple, with no value to society.
They probably look like this: □ □ □ □ □ □ …
YMMV. I’m sure a lot of people were shoving these things into backpacks, and this was post-2000, so the quality of disks for sale was poorer than it was in the 90s and 80s.
I’ve also (since then, for myself) repaired drives. The drives fail because they’re mechanical and have motors, and anything that moves or has a motor in it will need some kind of service. Servicing a floppy drive requires specialized equipment which is hard to find (or requires a lot of patience—it’s possible to repair a floppy drive with just a screwdriver, but it’s not fun; guess how I know?)
Magnetic degradation, wear from read/write heads, and environmental sensitivity mean disks have a natural expiration date and it's not long. And they are also easy to corrupt from all kinds of magnetic sources around us.
I get other stuff, even DOS like system. Safety critical software/hardware that was battle tested makes sense. Not floppy. Sometime it's just bureaucracy.
In a perfect world, we would move at a steady pace, as to keep up with the world but without rushing, and people would be accountable without a heavy bureaucratic processes.
Unfortunately, that's not how it works in practice, at least not for long, so between the two, we usually favor the slow and bureaucratic process when safety and reliability matter.
You can argue thats not a sufficient third redundancy, I don't have enough context to throw an opinion out there on the safety of this particular infrastructure project.
I am confident that the fallback process was considered and kept in place as the tertiary backup.
Excellent point though. It took me some thinking to put myself back in SRM mindset and what they'd have considered here.
There were two independent systems.
If this is the case, why was an airport knocked offline due to a cut fiber optic? A safety focused system would have a backup system say satellite or mobile 4G/5G. But they didn't have such.
This isn't about safety. Its about bureaucracy wrapped up as safety.
I would urge you not to consider this modernization from the perspective you may be familiar, as someone who may work on the bleeding edge and sometimes sees a lack of caution from engineers pushing for change.
The FAAs safety risk management is built around a rigorous process, but that is also a massive slowdown to reasonable progress.
One of the eng I worked with was 80+ years old and just didnt want to retire. He did not understand modern systems at all and the risks we were looking at were not (imo) realistically balanced.
I get the idea that the current administration is pushing safety to the side, but there is another risk where safety becomes a blocker to reasonable change.
When I left that work (more than 15 years ago) it was clearly a blocker.
If the last one is true, devs take note. If FAA can rely on slow, small storage for safety critical applications, you probably can reduce the resource requirements for your app and improve your customers’ experience and extend the life of their hardware.
I think this narrative is just meant to distract people from questioning the necessity of the landlord class, they want people angry at a faceless corporate boogieman instead of local landlords (for obvious reasons)
Basically when somebody makes a pricing error on a decentralized cryptocurrency exchange there's a mob of automated arbitrage bots that try to trade against it. Who wins and actually gets to make the trade is mainly determined by how high of a transaction fee your bot is willing to pay.
You can win instantly by offering to trade with a transaction fee so high it almost wipes out all but a cent or two of the entire arbitrage profits, because nobody else will pay a higher fee.
Instead what happens is bots offer the lowest possible fee, then immediately increase their fee as soon as they see a competing bot try to take the trade with a higher fee. This results in the bot offering the highest fee changing every several milliseconds as the fee quickly grows.
The way Ethereum mining works, this results in the winning bot being randomly selected from among the set of bots "bidding" up the transaction fee: blocks are mined at random intervals of iirc about 20 seconds, and whoever had the highest bid at the time the next block is mined wins. <20 seconds of bidding is short enough that even with the bidding, the transaction fee doesn't get bid up particularly high, and the winner makes out with a highly profitable arbitrage trade.
By giving up a guaranteed win and instead rolling the dice against other bots, transaction fees stay low and everyone has a higher expected profit. It's coordination that happens without any communication between the participants and quite possibly the participants didn't realize they were in effect cooperating when they added the high speed bidding behavior to their bots.
However, that's not a systematic method for lower prices. It just prevents individual cases of price gouging.
Okay but Realpage has a decade of real pricing data that it can use to continue informing these rental companies and outside of them having extremely hard boundaries in their service this is just going to happen again.
This doesn’t do a whole lot to address the actual issue, and these companies aren’t going to stop colluding over a wrist slap like this.
It’s just greediness unhinged by lack of policies. And the voters have finally had a say.
> Data center upstart Crusoe is being pushed aside at an artificial intelligence campus it was developing in Wyoming after failing to lock in customers including Alphabet Inc.’s Google, according to people familiar with the situation. https://www.bloomberg.com/news/articles/2026-11-06/google-ta...
Problems with Crusoe were already brewing back in June:
> The developer [Crusoe] known for giant AI buildouts announced that it had “paused” its plan to build a data center that would draw enough electricity to power a city the size of Denver. In reality, Crusoe is being pressured to leave the project, after Google had raised concerns about the costs and timetable under Crusoe’s watch, some of the people said, asking not to be identified because the information isn’t public. https://www.bloomberg.com/news/articles/2026-06-11/crusoe-pu...
Crusoe was previously a bitcoin mining operation that pivoted to building AI data centers around 2024. They don't have decades of experience building data centers.
It seems Google disagreed with how Crusoe was running the project and decided it would be better to own everything itself, so it acquired the project.
The map shown of 14th century Europe is doing better than most in that recognizes that several borders (but still too few) are fuzzy rather than lines on a map. The other problem you run into is that, in most non-state societies (in the Weberian sense of "monopoly on the legitimate use of force"), political linkages tend to be fragile and based more on personal ties so that a Big Man's successor may well repudiate the alliances and homages the Big Man makes. Which makes it hard to draw maps, because is Burgundy a part of France? The Duke of Burgundy is as often as not fighting the French king.
This is what makes adjudicating the "oldest continuous republic" hard, because at what point do we say that the homages San Marino is paying stops counting as being part of that country?
In reality, of course, they produce a great deal more than cuckoo clocks; one of the most competitive pharmaceutical industries in the world, industrial chemicals, defence technology, agriculture, financial services, food processing, etc - up to a trillion or so per annum IIRC.
If you're just talking about convincing a few individuals to do experiments, then there are plenty of people who fall for scams everyday.
Let's imagine that everything you say is 100% correct. LLMs can't access materials for developing weapons without weapons tests conducted at their behest, and humans choose not to - of their own volition - use world-ending weapons.
Let's also imagine for the purpose of conversation that one of the things that can be gotten at an extremely high skill level through increased intelligence is the power of persuasion.
A "superintelligence" persuades humans to perform weapons tests and then to give it the ability to use the weapon.
I think everything you've said is reasonable except the bit where you think step one of an "AI" exterminating humanity wouldn't look like a human doing it. In my opinion, the most likely vector for disaster is a human being who is "guided" by a computer model to do something unfortunate, not the model doing so directly.
But that’s just the parts of ML research that take place within your compute environment, namely experiments on model architecture, training strategy, UI/UX, etc. The algorithms and software.
Just because that part of research is progressively accelerated does not imply that we’re on a collision course with world-ending “superintelligence”. It just means that you’ll get to the best possible model of your data faster. And even when your data is the entire corpus of textual knowledge, there is a limit to the utility that provides.
It just means that software engineering, data analysis, and other computational tasks will be less of a bottleneck in the overall progress of the world (which, by the way, is great, and extremely valuable).
I don’t think it’s a stretch to say that we’ll have reached max LLM capability within a couple of years (what the author refers to as “super intelligence”). But it is a huge logical leap to say this:
> If we go through this feedback loop, I expect that the AI systems that come out the other side will be able to very quickly develop novel weapons and very easily overpower human civilization in a conflict.
First of all, the weapons necessary to overpower human civilization exist. They aren’t used as such due to geopolitics and humans’ basic survival instinct. Second, developing novel weapons requires physical tests, which are decidedly not accelerated by LLMs.
AI is a tool wielded by humans. If we’re wiped out by AI, it will be at the direction of a human. But we’ve made it pretty far with some pretty unstable humans in command of planet-killer weapons. I’m not overly concerned.
1. A key insight of foundation models is that capabilities are correlated. For example, training on code improved general reasoning capabilities: https://arxiv.org/abs/2309.16298
2. A profound capabilities explosion in easily measured domains will still feel very different to today. Like, the moment we come up with a systematic, objective way to verify a task, we become obsolete at it?
Benchmarks are by definition a curated set of things that can be measured and evaluated.
Things that can be measured and evaluated can be (expensively) trained on.
The progress we are observing comes from figuring out how to measure and evaluate particular aspects of intelligence, and then expensively training on feedback loops made of those exact measurements, and synthesized data representing further measurements (literally benchmaxing).
What is not shown in these graphs are the set of things that intelligence can do, which are difficult, expensive or impossible to measure and evaluate.
That set of things is much larger than the set of benchmarks represented in the graph.
Therefore the post is misleading in the sense that the "superintelligence" being targeted is "jagged" to use Karpathy's term, and we infact already have jagged superintelligence (calculators, compilers, Bayesian spam classifiers, etc.)
We are witnessing a profound explosion of much more jagged superintelligence, restricted to the exact domains that can be (expensively) RL'ed on.
This sounds truly Kafkaesque. And thank you for finally allowing me to use that word in earnest.
The other place they are amazing is in working with legacy code bases. I mean, the kind the dev team always wants to throw away. First, even if you do want to throw it away, you can first find out what it does and realize it does more than you think it does. You can also create integrations to more modern components which can be used to extend it. You can also create a pretty decent test suite if you do not have one. And, once the behaviour has good test coverage, you can re-write it into something more modern. You can even completely change the architecture while ensuring compatible external behaviour. Remember, this is all in place of “throw it away and start over from scratch, only to discover that the scope is much larger than expected”.
I’m also required to explain how I use AI in my work. I must plan all of my future work, with an AI tool. Another AI assesses whether my work is aligning with planned work.
Before I work and after I work, there is a million circular references that exist solely because of AI. These have scaled because it’s so easy to do with AI, but the we’re wasting so much time doing a ton of extra things very efficiently.
Let me do my job. My manager should know how I’m doing. Maybe highlight key wins. CSATs/etc go out every year.
If I had to do my individual tasks without AI, I would be slower. But if I take a step back, I'm really not changing the world any faster than I was before.
1. Intelligence per dollar spent/token output seems to be increasing rapidly. Opus 4 was $15/75 per million input/output tokens when released. Opus 5.5, which is significantly better, is $4/15. Sure it cost more money upfront to train Opus 5.5, or the bigger model (Fable 5.5) which tuned it, but once it is reached, there are massive diffusive effects.
2. The capital equation is still favoring more spend. Anthropic's revenue is growing rapidly and they are profitable on inference. Huge upfront spend for realized gains later on is standard for R&D. As long as serving the models is profitable and generates more value than the energy/compute consumed
3. The search space is too large to assume that improvement in AI assisted research is just models dumbly trying every combination until something works. For every additional bit searched, the space doubles, so even a trivially small combinatorial exploration incurs prohibitively large costs. Intelligent searching which continuously prunes that space along dimensions of effectiveness is not a trivial skill and is among the emergent accoutrements of these smarter models.
Anecdotally it seems like the majority of continued uplift comes from models doing more and exploring more of the search space, and we are getting better at giving them embarrassingly parallel problems to Hoover up with infinite token spend.
The other two formative experiences for me were using Cincom’s Mantis on the 3278 and Dataflex 2. Both made it obvious how quickly one could go from screen to persisted data, CRUD, application logic, and reporting with the right level of abstraction. Much later I described Dataflex as “Rails for the VT100”. Sadly, Cincom will not provide a hobby license to run it on an emulated 4381, and won’t even give away free manuals, and the company behind Dataflex (I don’t remember their name right now) seems to not even have any versions from that time. When I read Paul Graham’s Beating the Averages, it all clicked perfectly.