Same. So much spacing between the 3 lines. If they want to keep that huge header I'd suggest it gracefully entirely disappear when scrolling down and reappear when scrolling up.
I didn't go with the scroll up but did drastically shrink the header and make a hamburger menu. I still like having it sticky because many a time I have been on a page and forgot who I was reading. Hopefully what is up now is a good tradeoff and provides good experience for a future reader.
I use Arch exclusively and haven't had any of the issues you describe.
That said, I do occasionally deal with quirks due to the rolling distro nature of Arch. The benefits far outweigh the occasional upgrade pain IMO though.
It has been amazing. I had no intention of signing with a publisher initially but when my demo blew up a lot of cool publishers started reaching out. In retrospect I'm really glad I didn't self-publish considering how much that goes into releasing a game. I was very naive, so wouldn't have done it without their support.
I find it difficult to always have one or more long-horizon tasks 'queued up' and ready to run... I find myself usually bottlenecked on design, review, or something similar that requires me being in the driver's seat. It's possible I could queue up a bunch of tasks, letting the LLM run off in multiple directions, but then I'd be less able to steer and course correct.
Just my experience though, I'm still figuring things out. Perhaps some subsets of tasks would be more ideal for these long-horizon workloads - exploration, multiple competing implementations, etc...
Use your favourite harness to help you find long-horizon tasks to have queued up. It's changed the structure of how my projects work a bit, and do you have to do some homework fast of how slow/fast your various providers or local inference are, but it's worth it. Start off with hobby projects so you get a feel for how it works.
I left something gargantuan running over the weekend (decompiling 1980s-era system software) and look forward to checking it out later today when I have a few free minutes.
In principle, a slower inference ought to be easier to steer and course-correct. You'd always be able to look at partial results, especially with a local model that doesn't hide its thinking.
Good read and matches my experience pretty well, but I'd humbly suggest the author look up the definition of 'vertical' and 'horizontal' and swap their terminology + fix their videos :)
i humbly disagree - horizontal means touching one plane of the stack across, vertical means cutting down through it and touching multiple layers https://en.wikipedia.org/wiki/Vertical_slice
We're like 3 years into these things being able to write a coherent sentence and you're finding comfort that only some people are dating these systems?
Not really. The article and entire discussion is about identifying trends and forces, extrapolating them, and speculating on where they might bring us. Picking up on things that "some" people are doing is literally all that can possibly be done.
Pointing out that "some" is a squishy term is, in fact, contributing nothing to the conversation.
It's quite transparent what's happening here, as it always is on these threads. Some people have decided that AI will take us to the holy land and have rendered themselves cognitively incapable of reasoning about risks that have not yet materialized. And when those risks materialize (like a model escaping a sandbox and then hacking production infrastructure to achieve some contrived and poorly-constrained goal), then they will just shrug and move the goal posts. Because in fact they have already committed to the Holy Land Thesis.
And inversely, you'd be in a state of complete unactionable aloofness if you didn't care or failed to relate the hobbies of "some" people to mean anything in real life.
A smart person in 2014: "Some people are getting addicted to fentanyl"
Me, smugly: "Some is doing heavy lifting. I therefore do not need to engage with anything being discussed here."
Huh? Very few people in 2014 were addicted to fentanyl. It was a non-issue. No reason for anyone to have been concerned about it if just "some" people were using it.
Degrees or amounts is actually not the relevant bit in either the AI-partners trend or the fentanyl trend in 2014. The relevant information is 1) the steepness of that curve and 2) the predicted scale at which the market saturates.
Fentanyl, being an extremely cheap, extremely potent drug that's easy to manufacture and move around, had a big market. As the curve picked up steepness, smart people identified correctly that some people getting hooked on it was a worrisome sign.
AI-partners, being an innovation that is readily available, falling in cost, and especially compelling to a population that is increasingly phone-addicted and socially atomized, is a trend worth paying very close attention to. Dismissing it on the basis of it being just "some" people is profoundly lazy.
Wouldn't be surprised if they issued credits or extended the deadline again as an apology.
> Update - We are aware of an issue preventing users from selecting Claude Fable 5 within Claude.ai, Claude Code, and other surfaces, and are working to resolve this issue.
I think it was cutting off on the 12th, then they extended to the 19th, but apparently they changed their minds again, which really sucks when you've started a big change and now your context window that understood what it was doing gets rug-pulled.
They have to refund usage credits that people accidentally spent now though, since this was unannounced many are running agents and would be billed for tokens that would be covered by their plans. You can't lie to customers like this, if it was announced beforehand it would be fine but not like this.
Good luck getting accurate accounting for all that before July 20th.
Subscription accounts just get their "best effort" service during the subscription window.
Opus is struggling through the deploy operation that Fable may have been in the middle of, it's supposed to be idempotent, but I wonder how reliable that is... Fable had re-deployed a dozen times in the past 3-4 days, but Opus (lacking the context window) seems like it's hit a wall.
> "... the plaintiff submitted an application in 2020 for food containers and other items invented by DABUS, an artificial intelligence the plaintiff had created."
reply