"Humans have a token limit too" - that's so good and it explains so much of the fatigue that myself and colleagues/peers have about Claude in particular.
I think it's not just token limits - I think it's because it's so _dense_.
You get a week of research and debugging and testing compressed into a few pages. Even if it's explained well, it's just so much information.
And since it's AI, I'm constantly second guessing "is that really true?" and it's exhausting.
The scoring system is complicated but it comes down to passing about 4 new ones per year and then passing the renewal exams for all my current certs. Each exam takes about a month for me to be ready, maybe a total of 40 or so hrs. I have a family and my day job is very demanding so not a lot of free time.
There’s about 60-70 available certifications but they have various prerequisite certs so you can’t just go down the list. I was fortunate in that when the bonus structure first came out I had a lot of experience but not many certs so it was easy at first but now I’m running out of certs to get that are related to my actual day job.
However, I get the feeling the cert bonus thing is going to be cancelled next FY anyway. Sucks but at least I maxed it every year it was available and I have probably 75% of college for my kids paid for.
I've been using the Chrome CDP skill for Claude Code with great success for test automation purposes (locator detection, troubleshooting mobile layout, and so on). I found it here on HN: https://github.com/pasky/chrome-cdp-skill
I remember seeing another Chromium-based "MCP-focused" browser at that point in time, but I can't remember what it was called.
KidScreen is not an offline or download solution like USB or Plex; it streams public videos through the embedded YouTube player. Its narrower purpose is to make the parent’s approved list the child’s default shelf.
I recently migrated to a new phone and one of the apps I had to re-login was Dropbox. A feeling of nostalgia washed over me. It was absolutely essential many years ago but now the competition is huge. I'm not a fan of Private Equity but if it can keep the business alive, then it's worth exploring by the leadership.
Exactly ... PE's wants shorterm return, just numbers in a sheetbook, and they special execs for such jobs. Their only "strategy" (my dog can figure out better) is aggresive monetization, layoffs pushing remaining employees to limits.
Profits goes up ... so the strategy must work, lets increase monetization more and do more layoffs.
Prices goes up, quality of service goes down.
After few years, everything goes down the drain, PE and their execs can't figure out what went wrong.
In most cases you have the causal effect of PE incorrect. By the time PE comes in, the company has already peaked and is on a downslope. The original owners know this, want to get out and sell. Blaming PE is like blaming vultures for the roadkill they are picking over.
And despite their patterns giving not that great returns they keep doing it. Mostly it's an income play vs capital appreciation anyway. If you wanted the capital gains pre-seed and seed are doing well. ACA had a report which I think also got mentioned in Bloomberg.
That's an oversimplification. Private Equity's goal is to take control of the business and change it so that it has higher resell value.
Often they do sell it for parts, or they enshitify the hell out of it to squeeze revenue from loyal customers. Not necessarily because it's the optimal strategy, but because truly fixing a business is hard and these are decent shortcuts from their perspective.
But that's not a given, sometimes they do truly turn it around for the better.
I understand why it is a reasonable acquisition target for PE and worthy of exploration, but I’m not sure I understand how being acquired by PE would keep Dropbox alive longer other than shaking up the leadership and shifting priorities. Perhaps aggressively targeting enterprise customers currently using Box after rolling out the necessary features?
No, definitely agree. One of the great SaaS companies, strategy matters, though, and the early success distracted them from building a defensible business.
I was at Dropbox from 2016-2020. We were certainly trying to build a sustainable business, but there was a major identity crisis. Were we consumer web? Buy Mailbox and build Carousel, then shut them both down.
Maybe we’re Notion/Evernote? Buy Hackpad, plow a ton of money into Paper (which was legitimately good), then quietly deprioritize it.
Maybe we’re actually some kind of enterprise document productivity suite? Buy HelloSign. Plow a bunch of money into a desktop app. Pull more plugs.
A lot of smart people were trying. We made a lot of bets (too many?). None of them proved to be a second act, and the competitors eventually caught up.
I intend to write a piece going deeper into the failures; I would love to chat if you're open to it. Also, not to say Dropbox sucked or anything, it is just that the broader strategy and industry structure make it hard; if anything, the success of the first product made it difficult to evolve the business.
I've worked on a LLM-based visual validation framework for mobile layouts of different websites. The results are pretty flaky, even with Claude models. I've actually seen better performance with older ones like Meta Llama 4 or Amazon's Nova (the solution was running on AWS Bedrock).
I've since refactored it to first do some deterministic checks using CSS, to determine visual glitches like overlaps. Only after do I use LLMs to figure out if text is broken, loading spinners still display, or other such things where models are better.
For deterministic tasks like pixel-by-pixel comparison, I don't think LLMs are good enough (for now, at least).
I've done VHS (terminal recording) rewrite in rust built on libghostty called Betamax [1], and then threw it at creating a TUI for jujutsu. This worked pretty well, but is admittedly a much simpler task than a web ui. The harness was able to build out a feature, confirm it worked, find bugs in what it expected vs saw and fix them with evidence. Web is many more pixels and more nuanced things that have to do with how various elements relate.
I speculate that maybe the OpenAI models have the edge here over Claude's (while still being terrible at web UIs), as the training necessary to handle computer use stuff is heavily image processing based. I wouldn't bet on either being bad for much longer though.
It might help to consider how stable an arbitrary human would be at this task.
User interfaces are like pieces of artwork. You really don't know until it's finished, and even then it's not for everyone.
LLMs are most effective when working with human parsed representations of websites. They are terrible at doing the actual parsing.
They're also good at working on a relative basis. One png screenshot of the viewport is a difficult hill to climb. The difference between two screenshots is much more informative. Deltas usually tell a much bigger story than dead reckoning against the base world model.
This, but I'd say that there is an engineering perspective that you can apply. What the developer world is trying to discover here is how to encode taste concretely - often this is things like really mechanical design system things (space/padding/margins/relative sizes/color contrast/...).
The hard point is doing that well deterministically in a way that is token efficient and which can be iterated on. I suspect it's difficult to encode that into an AI model of what's good taste.
I am a Hover user for my personal website, on a .com domain. I've been extremely way of using any of the customized TLDs since I heard about these sort of massive hikes. I agree that the communication could have been much better on Hover side, especially with such a huge increase.
Really nice presentation and a ton of useful info. It's hard to boil it down to a single reading though. One thing that would be nice is a bit of a comparison between what companies claimed to do vs what they're actually doing, like a sort of BS meter.
Raising kids, however, can be very challenging in all sorts of ways. Physically, mentally, socially, etc. I became aware of just how much sleep deprivation affected me and for the sake of myself & my family I just sacrificed everything else to ensure I got good quality sleep. Fortunately, I was helped in this and I made damn sure everyone was supported when I was awake.
This also annoyed me a lot. I was running it before upgrading the SSD storage and I wanted to compare with LM Studio. Figured it would be good to have both interfaces use the same models downloaded from HF.
Had to go down the same rabbit hole of finding where things are, how they're sorted/separated/etc. It was unnecessarily painful
