The supervisor agent will keep the session in a loop until 6 hours have passed and eventually the agent will decide to use up the remaining time rather than fighting with it
I wonder how much co2 is being thrown into the atmosphere everyday through the steady stream of “look what I made this LLM do” and endless “benchmarking”?
Intuitively I'd say that just-for-fun AI projects emit orders of magnitude less CO2 than what just-for-fun road trips or air travel emit.
So unless you can show that it approaches all the other frivolous consumption that humans love to waste resources on, maybe we can get back to experimenting with cool technology and talking about it, on this website called "Hacker News"?
One useful hint here is the API token cost - in this case "about $74 in API tokens" for Opus 5.5.
We know Anthropic run inference at a margin, so that $74 means that the cost of the electricity involved is substantially less than $74. I'd love to know the actual cost there.
Approximately nothing - the carbon cost of electricity is already very low, and most of the inference costs are GPU and datacenter amortization. You could require all AI inference to be carbon-neutral and that'd barely raise the API prices.
I wonder how much co2 you use to get through the day, and how much you use to remark on other people's creations in a way that suggests you disapprove of their utilization of their available resources. How much co2 do your projects emit, since you seem to be interested in those metrics?
I strongly suspect one HN post (or two now) is not even in the same ballpark. I could ask Claude to figure it out but that would be a complete waste of time and energy. ;)
Make this the new benchmark please. The prompt is funny, the result is even funnier. I think the great part is there isn’t even a right or subjective way to draw it.
It’s frankly because Motorsport software guys are pretty arrogant. They look down on road car and truck engineers, but reality is they develop stuff with virtually no processes and they usually have only very basic engineering skills, yet every day they are told Motorsport is at the forefront of technology.
“Motorsport grade” auto components usually means crappy, and everything requires a software patch or constant maintenance to work correctly.
Not surprised there was a bug, and not surprised they were able to develop and deploy a patch in 1 hour without V&V testing it or anything.
I wouldn't be surprised if many of the motorsport software engineers you mention come from mechatronic/electronic/aero backgrounds. Don't get me wrong, I've seen an analog electronic engineer perform software wonders; 10,000+ lines of bash, custom C macro libraries, and a PHP runtime older than I am have taught me that programming can be closer to art than science. But relying on that cursed bundle of chaos felt deeply violating and unsafe lol.
Nowadays, I mentally place any engineer without production software experience into a 'research' box. Any resulting code may break the laws of physics, and should be tested as such ;).
Edit: It's also a minor crime to look down on automotive software engineering! The stakes of mass-manufacturing have forged some seriously awesome software. Top-of-mind for me is CAN support in Linux and safety-critical contributions going into ROS2.
Similar HFT and quants, they are elite in a very specialized niche: aerodynamics and CFD. The rest is extremely outdated.
Keep in mind that when James Volwes become the team principal of Williams in 2023, he was horrified to realize Williams operated everything in Excel spreadsheet, and to track the progress of how parts are made, they emailed Excel spreadsheet to each other (and remember, we are talking about a car with 100k parts). And the first thing he did was to onboard Atlassian / Jira and redesigned the process of communication / car production from the basics.
I have a friend who has worked in aerodynamics at several of the F1 teams. The parts themselves are pretty high tech, but all of the management of the data, the processes, the software to produce them etc is pretty shoddy.
Preach; I worked with a couple guys who left the F1 software industry and my experience matches yours. They wrote JavaScript like it was old school C, didn’t know version control etc, and refused to learn anything different.
Every time we fit out an office building we cover it 100% in cameras, both internally and externally.
Fact is, stuff happens. For example: we once had a crazed person run inside and start yelling profanities at people for no reason. Then, he denied it and reported us to the police for assaulting him. He got his girlfriend to act as a fake witness.
Having the cameras was a no brainer for legally protecting my staff, we could just pull the footage and show that no one touched him.
Similar protections for when male and female staff are working together alone. They are both protected from false accusations, one from harming the other, and the other from potentially false reports/blackmail that they attempted to do something.
We’ve caught interns stealing petty cash from people’s wallets (before cameras - caught in the act)
Is it sad? Yes. But it’s the world we live in and cameras are cheap. No such smaller incidents since the cameras went up. People behave better when they’re being watched, or at least they think we’re being watched. Nobody has actually watched the footage for months and it rolls over in about a week.
Referrer-Policy shows it can work. When the header is missing, browsers fall back to strict-origin-when-cross-origin. 86.6% of the sites we scanned don't send it, and we didn't count that as a failure for that reason. The other headers don't have a safe default like that yet.
I understand the need to make money, but hardly any open source software is free anymore even if you host it yourself. And this one doesn’t list the price which feels kinda cagey (get a license links to book a sales call).
It’s nice that it’s open source, thank you. I just wish that Mattermost et al didn’t open Pandora’s box with this one.
The supervisor agent will keep the session in a loop until 6 hours have passed and eventually the agent will decide to use up the remaining time rather than fighting with it