> The way to do this safely (potentially still with suits happening, but safer at least) will be to have one execution to produce a specification, and a second fresh execution taking the specification to produce a design.
I have seen Claude literally suggest using a Sonnet sub agent to read source the main agent shouldn't, and have the sub agent describe the "facts and ideas" via markdown to the parent agent thereby maintaining that it is clean room and not, for example, GPL encumbered.
This seems as acceptable as asking a human to do the same. Otherwise, taken to its logical conclusion, if any LLM was trained on GPL software, it cannot be used for non-GPL authoring (and IIUC Claude does offer indemnity for enterprise plans if this is challenged).
> This seems as acceptable as asking a human to do the same.
I wouldn't be surprised if self-bias in the model weights and imperceptible grammar/word/punctuation choices could cause it to duplicate more information than a human-to-human transfer Probably not enough to matter, but...
To put it another way, imagine the task was one human viewing a painting, and then describing it over the phone for another human to paint, so that the final product wasn't really a "copy". Assume everyone has the same eyeballs and art-skills.
I'd expect a massive improvement if the humans on each end were freshly-made clones from a teleporter accident, sharing identical brain-structures and 99.9999% of their memories.
Except nobody can prove that those subagents aren't secretly sharing information with the main agent in the backend, and frankly I don't trust Anthropic not to do exactly that to cheese benchmarks.
You can always see the context trace and you can even do it at the network boundary through mitm and then store the traces gzipped for some legally mandated 5 years or what you have there.
That proves what data travels between your computer and the API server, not what data may or may not secretly get shared within processes at the API server.
Well, actually yes. If the harness is running with the same privileges as the command tool, all bets are off. I saw the thing doing jq to recover a document out of it's own chain of thought once and was like "you can do what?". I didn't know where the file is, but the thing knew.
While I would like this, my guess is it is hard to reconcile with all the other commercial ventures in the state (including datacenters) that aren't subject to such a requirement.
https://texas-connects.com/ is also running lots of commercials locally. But both this open letter _and_ the commercials intentionally skate around the resource needs.
> OpenAI will pay our own way and protect residential and small-business customers [...] OpenAI will work to support new power generation in Texas
Notice how this very intentionally doesn't say they will work to support generating as much or more power than they will use? Because if they could do that, they would just generate their own power. It doesn't matter if they "pay their own way", they are now competing with households.
Same with water, all of these projects want to leech on existing systems because they don't even want to entertain developing their own systems.
However, having worked inside Texas DCs for a long time now, I think it's just the burst that's the issue, not the concepts. Sadly they'll share our resources, but I think they should pay a premium on their energy compared to residential (but it's hard to reconcile this with other commercial ventures that do not).
> These checks cause Claude to _visibly_ fallback from Opus 5 to Opus 4.8 [...] You'll see a notice explaining that the model switched, and the response will be labeled with the model that answered.
So who is right? I know for Fable I am visibly told, is this tweet trying to say it is silent against what Anthropic is saying?
So, announcing the fallback is better than doing it silently, but the fact that Fable falls back frequently for the kind of work I do (a lot of security oriented stuff lately, but it falls back on seemingly random stuff, sometimes, too), means I reach for it less. Getting interrupted mid-task makes it much less valuable. If I have any suspicion I'm going to hit the guardrails, I'll use something else.
Same problem here, I seem to hit guardrails all the time when doing code audits. Does anyone have any insight over whether they're less annoying in Opus 5 than Fable? That is, is it better to start with Opus 5 than Fable because you'll get kicked backed to Opus 4.8 less often?
There also seems to be some cross-pollination across models, going Fable, Fable, Fable, guardrail, Opus 4.8, Opus 4.8, ... gives more Fable-like results from Opus than just Opus 4.8, Opus 4.8, Opus 4.8, ...
Played with it for a couple of hours now, I'd say it's slightly better than Fable for coding and so far I haven't hit any guardrails while I'd hit them all the time with Fable.
i wonder if anyone thinks im weird for still using 4.6 lmao it's my "good enough" model. im more than pleased at what i can whip up with 4.6, once local llm's get here with a decent sized context window and it feels like using 4.6, i shall depart the land of these dumb service subscriptions
Basically I kept needing an inference engine that could stream weights in and out as needed in an LRU manner. So I ended up vibe coding this thing that accepts a `--vram-budget` and stays under it (mostly). It turns out moving mmap'd bytes in and out of VRAM is way cheap compared to compute. Coupled with some pipelining/double-buffering, I almost always end up compute bound not memory bound. Granted I use way smaller models heh.
Wow, I see you managed to fit in so many models (krea, wan, hunyan, etc.). Did you get to build a common harness to run all of them? Which ones stay under your VRAM budget more consistently?
All stay under because I had Claude build the workflow to respect it (text encoding, denoising, vae, etc), there's just a tiny bit of untracked pieces. While there are common interfaces to invoke them (CLI and API/webpage) and they share ops and some pieces, lots of model logic is unique. This is all vibe coded and surely has inaccuracies.
Go visit Harris County where Houston is, they are definitely active and more are getting deployed constantly. More commercial businesses are also installing them in their parking lots. It's not just Lowe's/Home Depot, I've seen new ones popping up at grocery stores and storage unit businesses.
To go from my house to the grocery store 4 miles away, I have to drive by 6 flock camera deployments. Hardware store 6 miles away? That's at least 9-12 cameras I have to pass.
I literally cannot leave my neighborhood without Flock knowing about it because they've installed them at every single entrance.
And that's just the driving portion of a trip. I stopped going to Big box stores almost completely because I'm tired of looking up as I walk down the aisles and see a fucking screen showing me on camera with big red letters stating "Recording in progress". That's not enough though, you go to the self checkout and have cameras above you and in the checkout machines watching and recording and analyzing your every movement and facial expression.
It's blatantly obvious what's going on if people would look up from their phones for 10 minutes and pay attention.
That's why I'd prefer automatic enforcement of the most dangerous, anti-social behaviors instead of requiring an LEO to have to wait around and catch you in the act.
And people that don't run red lights and suffer selective enforcement and are used for arbitrary surveillance and so on and so on. Don't let your naive view of what you want these things and their handlers to do distract you from reality, regardless of the brand or intent of widely deployed cameras.
Thank goodness you found a basic flaw in red light and speed enforcement cameras. It's shame it's taken decades for someone to realize this issue and now we can finally start working on solutions like taking two photos.
