News for hackers. Stuff that matters.

Moto7451

This is mostly a memory bandwidth issue. Qwen 3.8 27b and the related Bonsai 2 27B run on my M5 Max laptop with 36GB of ram. It’s a little slow due to my specific model having lower memory bandwidth than higher end models with more ram and cores. MoE models perform better here and expert streaming engines can allow larger than memory models to run.
From experience, if you work somewhere with very disconnected higher ups a sales person shows what your token usage is vs spend on the monthly subs for whatever set of users you have and then show an API price with a sample bill for even “more users” at that same number of tokens that is 30% lower. The crux of course being that your existing user base is not going to use fewer toke a just because you signed more people up and have a discount for the status quo.

Then when you consolidate all users to a single provider you find out you have a 2x to 3x larger bill because you… logically have more users at the same average token spend than the quote.

At my previous job, all while managing a full team, I did a lot of porting of legacy apps to modern standards to maintain various obligations. It was AI gen work in between meetings and paperwork. It was seen as such a success that I was asked to move to an IC role to focus on it within the “Elite new products team” that was supposed to be all AI and the latest tech. I showed up to find that they were pushing out low quality manually written code in 12 hour days six days per week to appear like LLMs were in use to save face because they lacked a token budget or the skills to properly manage one under the APIs. They had dismissed someone who had spent multiple thousands per month on Agent swarms. It was the perfect situation to make their 3x budget issue a 10-30x budget spend issue.

For many many reasons I left that job and got a new one but that was the cherry on top.

Let’s separate out what the companies actually want from the words they’re using for a moment. The top commercial companies are roughly stating:

1. Our AIs are so advanced they’re busting out of our networks and hacking people. We can’t seem to stop them.

2. There are real dangers to just how advanced our AIs are advancing.

3. Someone needs to slow us down

If these were true statements then they’d likely want to shut down themselves for the sake of the planet they live on. If there was a real problem as they have stated then shutting down is the solve (and to be clear, I’m not saying there aren’t problems with the AI industry. I just think they’re BSing us).

In reality this is marketing nonsense (better get in before the ban!) and “our” AI is never open weight/Chinese AI models which they have a different lobbying arm against. If the top four companies got a loosely regulated monopoly through regulatory capture, they’d be giddy.

Re:EV Sales Are Booming in Europe with Gasoline at $10 a Gallon on Sunday September 27, 2026 @11:00PM (#49873080)
I’m in the US but I’m basically charging the same way. 16A 240v. I’ve accidentally stepped it down to 13A and didn’t really notice. If you sleep, and can plug your car in while you do so, you can charge.
Re:ReBarUEFI: Resizable BAR for almost any UEFI system on Tuesday September 22, 2026 @07:40PM (#49809773)
I think you’re experiencing paging to main memory rather than anything ReBAR related. ReBAR itself reduces memory copy operations. It’s of course possible you’re experiencing some sort of BIOS bug.
Re:ReBarUEFI: Resizable BAR for almost any UEFI system on Tuesday September 22, 2026 @07:39PM (#49809759)
Resizable Base Address Register. In short, rather than the 256MB of mapped memory available to the CPU for any PCI device, ReBAR capable devices can map larger memory to the CPU’s addressable memory space. Without ReBAR or sufficient tricks (that Intel GPUs lack) you have to shuffle 256MB chunks around between GPU and CPU.
Re:Apple Copland D11E4 Booting in the Browser on Monday September 21, 2026 @10:16PM (#49796039)
Mac OS 7 “Virtual Memory” was just the swap file. Fundamentally classic Mac OS had a flat memory layout with special hooks defined that did not allow for virtual memory as more commonly defined. Multiprocessing came in asymmetric and symmetric forms but really wasn’t terribly useful until 8.5/8.6 and later but by that time Mac OS X was about to show up which solved all these issues.
They can just add those people to the board. It’s not a set number and they don’t regularly add or remove people unless they think they can bring something to the table. Sometimes that is simply “holds a bajillion shares” but if they wanted a High Frequency Trading CEO on the board for insight into that, they could add both his kid and that CEO.
Re:Trees use a "muscle", tension wood, to correct their posture on Wednesday September 09, 2026 @12:03PM (#49628744)
I have a fun entry to the exception to the topping rule. I have a diseased sycamore that was really struggling. The disease was primarily at the top of the tree. We were going to cut it down to prevent neighboring trees from being infected or the tree from falling. The cutting company had a “hold my beer” style suggestion that topping it would bring the tree back. Boy did it. It’s flourishing with new branches from the cut and beautiful coverage. The fungus hasn’t come back and I’m not worried about it dying and falling on my house. Everyone wins!
Re:GPT-5.6 Sol Pricing Cut by 50% on Wednesday August 19, 2026 @12:27PM (#49363686)
It's where the stuff ultimately goes since the creditors aren't interested in GPUs that no longer have much value. The context here is a crash, not just a basic bankruptcy. If it were that then absolutely it would be impounded by creditors (virtually or in reality) and then sold to the highest bidding data center.
Re:GPT-5.6 Sol Pricing Cut by 50% on Monday August 17, 2026 @08:57PM (#49339802)
I would in such a scenario expect the GPUs to be dumped to industrial breakers who would send them to China for refurbishment and repackaging before being sold again on Amazon, AliExpress, and Taobao as last gen gaming cards from weird brands and specs.

This is what happened after the great crypto GPU dumping.

Re:I Made Tinnitus My Friend (Then It Disappeared) on Monday August 10, 2026 @11:41AM (#49245130)
That anecdotally matches my experience thinking about it.

After my initial hearing damage from someone clearing a rifle before everyone put on hearing protection it took a while for it to attenuate to background noise but flares beyond that volume. Things that certainly avoid flares is when I have a good sleeping setup where I avoid being crumpled or my neck kinked. I have good posture in general but sometimes weird working locations, off site meetings, hackathons, etc would lead to my tinnitus getting worse.

I chalked that up to being physically tired from travel and lacking sleep (both of which seems to also increase my tinnitus)but when reflecting back I don’t recall having much Tinnitus when my son was a newborn through a one year old and I basically slept walk that whole year.

I think on the count back the neck posture and likely the muscles in my neck and head lacking tension is what is keeping it from being worse.

Precisely. As someone who has done work in my home that has passed inspection, there’s a lot more to it than just taking a test. It’s the “how things are supposed to work” meets “how am I going to actually make this work” X factor that takes time and ultimately what people pay for.
Maybe not as fun but my sun room window unit is controlled by an off the shelf $20 thermostat designed for Air Conditioners. It’s the LUX Win 100 and is date/time programmable and features a compressor lockout so it doesn’t burn out your window unit by short cycling.

If you have a favorite smart thermostat you could also wire up a receptacle with a contactor between it and a cord and a 24v AC transformer wired to the line side. This would make your window unit controllable just like a central air conditioner.

This is why you pay the premium for epoxy grout. When we redid our house we used only epoxy grout and cleaning it as as simple as using a foaming tile spray.
Re:Remind HN: Today is Mother's Day, call your moms on Sunday May 10, 2026 @01:42PM (#48086073)
Yes
Thermaltake already makes a clone:

https://www.tomshardware.com/features/noctua-nf-a12x25-vs-to...

Noctua seems fine so long as you’re not copying the color scheme and branding. Interestingly TT had a 140mm version before Noctua. Noctua seems happy being the premium option.

Re:M 7.4 earthquake – 100 km ENE of Miyako, Japan on Monday April 20, 2026 @01:58PM (#47838140)
Evangelion is not “Japan’s version” of anything in terms of western religions. It’s one guy who was making a TV show. It’s not Japan’s “national personification.”

If you want to go down the rabbit hole you’ll find a lot of quotes from Anno and others on it.

A lot of it boils down to “we did this because it’s cool and we have no idea what Christianity is” down to Anno going “Yeah I could do this because no one in Japan is really going to care and I don’t really care about the Western world anyway.”

https://gwern.net/otaku

https://wiki.evageeks.org/Statements_by_Evangelion_Staff

It’s fine to not like it but it’s quite a step to go “this is how Japan thinks.” It’s akin to saying America is personified by the Simpsons every time they go to Japan.

Re:What Young Workers Are Doing to AI-Proof Themselves on Monday March 23, 2026 @10:07AM (#47489757)
This is precisely my experience. Except our roadmaps didn’t change much.
Re:What Young Workers Are Doing to AI-Proof Themselves on Monday March 23, 2026 @09:56AM (#47489616)
No because you have the customers and budget in that moment for three already and I’m not interjecting that more budget is given or any taken away.

Restated, I’m not saying we’re hiring more or less because of LLM AI productivity changes. I’m rejecting the idea we need less people for all the previous reasons stated and my own two cents that we’re yet to see the reduced work hours Keynesian economics predicted as output per hour increased. We humans just keep working the same hours even if that hour is massively more productive.

This last point is well studied and not my own original thought. I’m just poorly regurgitating college level Macro Economics.

My own point I’ll add here is we’re not seeing companies bragging about their two day work weeks.

My personal experience with layoffs is that it’s all been financial engineering and the lack of nearly free financing that we had in the 2010s, again in the Pandemic, CapEx tax changes last year, and/or over hiring similar to but not nearly as massive as Google and Facebook. I worked for a European company that hired a dozen Americans to become more “US Tech Company” like and eventually let us all go two years ago once the fun money ran out when rates increased. They did a little bit of the AI babbling but realistically they couldn’t get the financing to keep it all rolling.

The companies reducing to one developer for a product are likely not doing this because of LLM AI work but likely will survive better because of it.

To your point, I’m actually living it and it’s nothing to do with AI. One of my teams was cut to 25% of its size a year ago and the whole QA team let go. Roughly this was an EBITA play. Basically the only way we get anything done is by doing what I mentioned in my earlier post where the front end dev uses LLMs to build a prototype backend they can use to support their front end expertise and the back end dev does the same for the front end. Eventually they meet in the middle and I can juggle some of the KTLO myself. Is this fun? Absolutely not. If we had the headcount back we’d be able to meet the ‘25-26 roadmap but instead we’re doing 40% of it.

Re:What Young Workers Are Doing to AI-Proof Themselves on Monday March 23, 2026 @12:56AM (#47485625)
Yes-Anding you, if one developer is suddenly 3x more productive and does 40 hours of work in 15 with AI tools, any reasonable manager would still want the same three people to keep working at 3x productivity.

I don’t see Keynes’ theory we would all be working drastically fewer hours per week suddenly materializing due to AI. As always we’re just going to try to output more in the same time. The fact I, a manager, can “vibe code” some bugs away between meetings does not mean I will benefit from having one less dedicated engineer.

Re:How often do full-body MRIs find cancer? on Sunday February 15, 2026 @02:02AM (#47021654)
PET Scans feature areas with blood flow so tumors show up as hot spots for follow up. People who are maybe only feeling off or had one confirmed tumor can have a lot of small tumors spread across their body which will show up clearly on a PET scan.

When my brother was at the end of his run fighting cancer he felt a bit under the weather and managed to catch covid so everyone figured he was feeling bad due to that. The PET scan showed he had thousands of small masses converging into the large mass that eventually killed him by cutting off blood flow to his kidneys. His cancer was an aggressive blood cancer that had stood up to conventional and Trial Chemo drugs. There was no way to treat this but other cancers that are less aggressive can be treated at this point and would be treated differently than a single mass.

In a nutshell it’s that level of visibility that makes PET scans worth it.

Re:I reverse-engineered Netflix's 4K restrictions on Wednesday February 11, 2026 @04:12PM (#46981047)
The same plugin spec runs across several browsers and Edge is now chrome based. It’s likely just hard coded defaults that seem a little silly when used on the target browser it pretends to be.
Re:Toyota Fluorite: "console-grade" Flutter game engine on Wednesday February 11, 2026 @04:10PM (#46981025)
Yup. The expectations are set higher and to a point since cars are bigger for safety reasons (crumple zones, airbags) and have more pedestrian safety features like spring loaded hoods, it invited incremental additions until the new price points were set. A spartan 19K car isn’t going to sell as well as a CarPlay equipped 23K car.
Re:Toyota Fluorite: "console-grade" Flutter game engine on Wednesday February 11, 2026 @01:17PM (#46978642)
A Toyota Corolla starts at $23K. I think the "Under 20" and "Under 30" price points (a la the original Model 3 goal) are simply a thing of the past for any volume car with reasonable demand.
Re:I reverse-engineered Netflix's 4K restrictions on Wednesday January 28, 2026 @07:34PM (#46803982)
I believe the benefit for Edge is faking HDCP 2.2.
Re:What Has Docker Become? on Sunday January 25, 2026 @02:59PM (#46757589)
Yup, exactly that. The situation shouldn’t have even happened in the first place but sometimes that’s just how it goes.

I’ve consulted at some places that should have been licensing Dicker Desktop and I just skipped the workflow headaches and used Rancher from the start with the compatibility aliases installed. A lot of places are simply unaware.

Re:What Has Docker Become? on Sunday January 25, 2026 @02:54PM (#46757549)
I don’t object to licensing software. It’s the way they went about it. Individual EMs and ICs being targeted and intimidated vs going through a procurement/legal channel. The companies I’ve worked for have staff lawyers for a reason. If they have a legal objection they should take it up with them. Not someone trying to work through a React bug.
Re:What Has Docker Become? on Friday January 23, 2026 @10:09AM (#46733409)
I believe you’re using royal-you but just to be clear I didn’t run these companies.

At one place there wasn’t and at the other it wasn’t well managed. I agree from a compliance point of view and have advocated for this but I was not on the IT/Ops side of the business so I could only use soft power.

The CTO at the first company had a “zero hindrances for the developers” mindset and the latter was reeling from being the merger of five different companies. The latter did a better job of trying to say the least but wasn’t great about it. Outcome was the same none the less.

Re:What Has Docker Become? on Friday January 23, 2026 @08:54AM (#46732506)
One thing that really hurt them from my PoV was how they acted when they changed their licensing structure with respect to revenue generating companies. I’m fine with the idea that licensing Docker and Docker Desktop is a good thing to do. However, I think they just made people distrust their motives with their approached to this.

At two places I worked their reps reached out to essentially ensnare the company in a sort of “gotcha” scheme where if we were running the version of Docker Desktop after the commercial licensing requirement change, they sent a 30 day notice to license the product or they’d sue. Due to the usual “mid size software company not micromanaging the developers” standard, we had a few people on a new enough version that it would trigger the new license terms and we were in violation. They didn’t seem to do much outreach other than threatening us.

So in each case we switched to Rancher Desktop.

The licensing cost wasn’t that high, but it was hard to take them in good faith after their approach.

Older »