>> Wait until modular nuclear

I'm waiting to see some successful pilot plants. 5 years or more?

>> makes way more sense than solar

No it doesn't.

You will be walking along looking up in the corner of your vision at the little display, and you will be making jerky motions with your arm to tell it what to do. All witnesses of this will feel sorry for you.

" Chief AI Officer Alexandr Wang's team has abandoned Meta's previous frontier model Behemoth and is developing a new model from scratch"

All the AI that Meta had done previously is third rate and therefore worthless. Zuck realized that and poached the best talent from the competition (at enormous cost but he's swimming in money). The new folks have reviewed that legacy code and wisely decided to start from scratch. I think it will take a while.

>> Most of that smog you see is tire particulate.

No it isn't.

>> we don’t have much to do with it

Obvious lie.

I'm a full stack developer and currently I'm working on a machine vision project. Its a large codebase and lately I've been breaking it out into a few much more useful microservices. Each one has its own little administrative webpage, a test suite, and a comprehensive readme. 90% of that new code was written by AI as a result of hundreds of incremental prompts. Looking at the AI dashboard I'm seeing that it wrote more than 11k lines of fully tested code for me over the past month.

Yesterday I was wanting a small bug fixed, the result of a previous refactoring. I pasted the error message into the console and pointed out the problem area. The discount AI I was using chewed on it for several minutes and eventually it proposed a bizarre major rewrite of the whole file. No way buddy! I switched to a better (and more expensive) model that identified the bug and fixed it in 30 seconds, about 5 lines of code. A few experiences like this and you figure out which ones you can rely on to do a decent job.

I've used all of the top 10 models for coding at one time or another and in my experience there's a significant difference in performance. The open source models aren't even worth bothering with. Some of the closed ones aren't either, even though they may be free or available at a large discount.

I can understand why companies in the West aren't willing to spend a huge amount of money and resources to compete in the open source space. How do you ever eventually get paid? How do you even attract users if you aren't (relatively) any good?

I suspect the Chinese government has made AI a national priority and they are providing the subsidies that fuel open source models there. And that AI is being made widely available to the populace for free, all part of the plan. DeepseekV3 is available to me for free, not interested.

>> Sounds like a manager

Not a manager at present, but if I'm doing the hiring I get to decide whether the candidates are going to be local and drive in or remote. I'm paying cost of your commute as part of your salary. I'm also paying for your room and board, and I don't like whiners.

>> It's comparing open to open, not open to closed.

That might be what they are comparing, but it isn't what I'm comparing. All of the open source models are far down in the rankings, easily beaten even by several of the free closed source models. I see no cause for alarm in Washington and Silicon Valley.

I've tried the Deepseek and Qwen models for coding tasks and they are nowhere near as good as the top-line competition.

Rankings here, the Chinese models aren't in the top 10;
https://ancillary-proxy.atarimworker.io?url=https%3A%2F%2Fllm-stats.com%2F

>> I'm fully remote only, and refuse to go into an office

Sounds like whining to me, Anonymous Coward.

I started using Bitbucket long ago when github wouldn't let me have private repositories without paying them money. That's changed now but I see no reason to go back.

"the crew studied space-caused mental and physical changes in astronauts, blood flow from the brain to the heart" and their eyes. Apparently a long interval of weightlessness screws you up royally and we are now trying to find out exactly how much.

AI has found and fixed quite a few bugs for me, but what I use is integrated with the IDE and can see the entire code context. Also it will write debug logging statements etc to zero in on problems and evaluate the results. It won't necessarily do everything for you but it will help.

But in this case it appears that their test case was looking for vulnerabilities "in various popular open source software" which is a whole different thing and presumably harder.