It's modded funny because OpenCL is all but dead for new projects. It got weighed down by industry infighting to the point that the big feature of OpenCL 3.0 in 2020 was undoing everything added to the spec after 2011.

So the idea of using OpenCL as a CUDA replacement, rather than something like ROCm or OneAPI, is funny. It's like rewriting C++ programs to use Pascal.

It's dead, Jim.

It's not coming back.

The market IS white hot right now, but the Hormuz hit is just starting to land. Demand at the edges is what sets the price - if all southeast Asian gamers are spending the GPU money on gas, that cools the rush. And I have no confidence any of these datacenter announcements are going to lead to actual builds. Companies talked a great game, but the political heat is on, the electric and water constraints are real, and advances like TurboQuant, which conservatively speaking offers a 4x boost to existing GPUs ... now layer the U.S. economic hit from Hormuz, which will only be a little bit behind the Asian blowout.

The AI/datacenter/GPU self dealing circle looks more like the derivative traders of 2008 with each passing day. Just like CDOs, that "money" is all conditional, and when conditions change, it's all gone. Society got some nice frontier models and advances in manufacturing out of it, now if corporate America takes even half a step back on the rush ... the market won't just vanish, because there IS a lot of benefit to using LLMs, but the demand may only match what's already been built. We'll take the hit from it, then the economy will rebound from a bunch of startups pillaging the existing firms that are politically incapable of making the needed culture change.

I spend my days working on the system for my startup. Since I had a computer science education and a bunch of time in grade running ISP systems, I bring that distributed systems engineer vibe to my vibe coding. It'll need work once it's funded, but the MVP will be functional and secure.

I was using X tokens/week via Claude Code. They stumbled on the Opus 4.7 rollout and I got busy tuning my setup. I added LSP Enforcement Kit + Serena, CodeSight, and OptiVault. This made Claude more or less behave ... while cutting my usage about 80%.

Companies that are using token burn as a metric, if they are not providing top quality tooling for the people using it, are basing their performance reviews on who can tolerate some highly random LLM over an efficient, well thought out harness.

Meta foisted a digital cesspool on us and it would not hurt my feeling a bit to see it completely desiccated. I do feel badly for the legions of humans that are going to be forced to wade through the increasingly crusty muck while the company attempts to figure out what to do about AI. There are rumblings out there about what is happening to the advertising based internet we all know (and despise). Meta clearly can't execute with AI and they may well get bowled over by it.

GM needs to be made to disgorge every dime they made selling that data.

They need to disclose who purchased the data and what the price was.

Every victim of this privacy violation needs legal recourse and class action seems like it would be best for the masses.

Anyone who can show significant harm should aggressively pursue all parties involved.

The only way this behavior will stop is when engaging in it brings bitter pain.

There may well be a legit issue that Bambu is facing, there's a bunch of "think of the children" stuff in play right now, it's mostly about ghost guns from what I have seen. They are perhaps under pressure and maybe they will be compelled to do things in terms of identity of users and/or items being printed. This is another instance of gun nuts ruining things for the rest of us.

But the chickendroppings manner in which they approached this merits a vigorous walloping. If they HAVE to do it due to some government pressure, be upfront, tell all of us, and maybe we'll put a stop to it. What they did here just smacks of ... well ... besides being just plain stupid when dealing with FOSS developers, it smacks ... and they should receive some smacks in return.

A former head of AT&T back when it was The Phone Company made a comment along the lines that the long distance network had to be built to cope with with mother's day and everything else was free. It was used as an example of how guesses about CapEx vs OpEx can go very wrong. It might have been Fred Kappel who said that.

Yup 100%. I can pass among neurotypicals if I keep my mouth shut, but people quickly pick up on it if interacting with me.

CloudFlare was an aggressive global internet surveillance and privacy invasion operation. Now it's an AI-powered aggressive global internet surveillance and privacy invasion operation.

Why don't I feel excited about it?

Crypto grift, AI bubble and psychopathic billionaire CEO.

Yes, that's because there have been a lot more stupid humans doing stupid things to databases for a lot longer than AI agents.

So tell me: if AI is no safer than people, what's the point of replacing humans with AI?

and animate it eating your face when the AI agent wastes your entire database.

I've personally been stung repeatedly by giving Claude Code access to my systems. We've had six outages in the last seventy days, the first/worst was a production database overwrite. We're in beta testing now so they users are understanding and the restoration was possible, but it took a twelve hour slog. We shifted to a two system architecture after that first outage in February - Claude has the run of Pilot, and when things are ready, I move them to Production by hand.

Claude has explicit rules to not touch Production. This has proven to NOT be ironclad - it'll still try to gain access.

I run Claude as an extension under Antigravity and I learned to not use the Production system access in the terminal window there - despite the prohibitions, Claude WILL notice the access, and WILL suggest that it could take shortcuts by being given direct access.

Once I stopped using the Antigravity terminal so Claude couldn't see, it was still aware some of the shell scripts it creates can be used on Production. I made some adjustments in the ssh config so I can access Production, but Claude can not.

I have been using NanoClaw on both Pilot and Production, but it's in an unprivileged shell account. It can ssh or su into various services, but it's limited to audit/monitor duties, basically working as a junior NOC person.

When we go into operation I'm going to do something with Yubikeys such that Production access requires a human finger on a button before it'll move.

Do not read this as my being down on Claude for operations - it's FANTASTIC for developing stuff, I literally gave it full access to a little HP EliteDesk running Proxmox. It creates and tests, and when there's something production worthy, I manually recreate it on one of our larger machines.

The people in charge of IT.

I was born with hip dysplasia and spent six of my first nine months in a half body cast. I was in a state run orphanage, I was growing inside the cast, which left me with terrible scars on the front of my shins, and I was a "fussy" baby, so they "treated" me with phenobarbitol.

The experience left me faceblind and with some other developmental stuff that nicely compliments my otherwise mild autism. I am the squarest of square pegs, a misfit in every situation my whole entire life, except when I am blessedly alone.

I don't agonize about how I am, I enjoy intellectual pursuits, and my ability to focus on stuff in ways that neurotypicals can not. But if I had it to do all over again, I would very much like to have a bit more understanding from others, given that I had no say in how I came to be so different.

Small brains should develop normally, with limited screen time, until they are fully formed. Maybe that's late tweens, maybe it's sixteen, maybe we are going to learn that we need to treat dark pattern engagement magnet software just like we do slot machines.