567 karma · joined December 7, 2024
If they offered 3 Flash (or 3.1 Flash Lite too but might be hoping for too much) with comparable usage limits then the transition to Antigravity CLI wouldn't have bothered me much at all.
I was going to say you should submit it but I saw you did a few days ago but it only got a few votes... If Dang sees this IMO it would be extremely deserving of the second chance pool as I wouldn't be surprised to see easily jump to the front page with a different roll of the dice.
Although realtime detection of changes on the file system is still a little flaky for me (possibly how I’m running it).
Google or anyone else could start adding those unique tracking watermarks you're concerned about any time they want, regardless of whether they use this AI detection watermark, that to be clear can not track you in any way.
> The result is that the illegal market dwarfs the legal market. The legal suppliers simply can't compete with efficient and untaxed illegal or grey market sellers.
Even with California taxes an 1/8th of flower is often less than half what I’d typically pay back in 2008 or so, even without adjusting for inflation.
Also, I suspect middle class buying preference in the last decade or so has heavily shifted to gummies/edibles and vape carts which are much sketchier in the black market vs relatively interchangeable flower.
The idea of smoking a literal bowl to get high wouldn’t even be in the first like 5 methods among the people I know. It’s not super appealing in your 30s-40s while living in apartments/not wanting to reek of weed in the office. So buying off the black market just isn’t attractive even if possible cheaper.
Within the framework of your analogy, it's like responding to someone active in DIY maker groups suddenly dealing with an influx of influencers in meetups showing off Chinese junk from Etsy to post on Tiktok, and accusing them of being a Luddite blinded by their zealous hatred of mass production -- both strangely abrasive and also fairly nonsensical except as a "mass production supporter" social signifier.
Not to mention, in the article they specifically describe themselves as a heavy user of frontier models for security research ever since the release of Opus 4.5, calling them "useful within the field". In fact I don't see any actual criticism of AI/LLMs anywhere whether for security research, programming or anything else, except for making competitive CTFs no longer viable.
What does it take to avoid the "Luddite" brand? Using AI themselves and praising AI as useful (to the point of having a lopsided advantage over humans) isn't enough? Do they also need to say "I haven't written a line of code in 6 months/it's easily a 100x multiplier for my job" every time they mention it too?
> Vietnam - unnecessary war, but we won the peace
> Afghanistan - we tried our best and made some mistakes along the way. Eventually got Bin Laden though. *Too bad the rest of the world didn’t help.* Now we’re seeing a massive regression in women’s rights there.
> At its peak between 2010 and 2012, ISAF had 400 military bases throughout Afghanistan (compared to 300 for the ANSF) and roughly 130,000 troops.[7] Forty-two countries contributed troops to ISAF, including all 30 members of NATO.
In fact it gives the opposite impression by specifying "at least one", which implies the category is supposed to be distributed widely enough to be recognizable in an org of sufficient size, but not dominating the ranks of software developers in droves. That seems more like a strawman you're arguing against.
Warning
Heads up: This is a research project — bounties listed here are symbolic and part of an academic study on open-source contribution patterns. PRs are reviewed for research purposes only and will not be merged into production. If you're looking for paid bounty work, this is not the right repo.
I guess a sort of task blindness where once they've gone as far as to git clone they've already switched gears from searching Github for qualifying bounties into a find bug->fix bug->open slop PR mindset to close the loop and end the turn? By that point an incidental warning they ingest in passing while looking for the Solana contract vulnerability they already committed to working on in a comment might not even register as relevant to the current task at hand.
She called me back later that night and we chatted for bit and then she paused and sort of uncertainly was like “So… was there something you were needing to tell me?” And I was completely baffled and was like “Uhhhh I don’t think so…?”
She then explained the notification she got about my call and apparently the LLM summary of my voicemail converted a message consisting of 75% well-meaning but insignificant interpersonal human filler (like most voicemails) into this stilted, overly formal business-y speak with a somewhat ominous tone. Assigning way too much significance to each of the individual statements in the message about wanting to talk (to say happy Mother’s Day), inquiring about her availability ASAP (to say happy Mother’s Day) etc. Plus grossly exaggerating the information density of the call making it sound like I left this rambling, detailed message about needing to tell her something that was left completely vague, but possibly important and also time critical.
Added up it made her a little worried when she read it and made me a bit pissed that was the end result of my wishing her well. Because apparently everything needs a half baked LLM summary crammed into it now.
Even if the coding process itself is “effortless” and the agent just churns away to implement whatever I ask for on a dime, it can become exhausting thinking through all my needs/wants, tradeoffs, API shape etc. Despite not needing to write a line of code myself or read more than excerpts in the chat it can turn into a slog after the honeymoon period passes and it starts to feel like unpaid work.
I’ve had moments where I’m relieved to discover a popular open source tool that works out-of-the-box as an alternative to my own so I can offload that organizational overhead and decision fatigue to someone else. While benefiting from all their features/enhancements I didn’t have to design or maintain myself over time.
As an example, I had been building a TUI/web app to download and organize ebooks from various sources like Project Gutenberg or Anna’s Archive with a central meta search, and manage my personal library. It solved the immediate problem at the time but I kept needing to add missing features, plug holes in the various search integrations, UI refinements, etc and it never quite worked exactly as I wanted so kept having to work on it and became less and less fun as time went on.
Then I discovered Calibre Web Automated + Shelfmark on GitHub that did 99% of what I needed plus a lot more and overall had a level of polish and reliability my tool never reached. Now I just pull a Docker container every so often for updates and made a few tweaks to syncing but overall spend vastly more time on actually reading/organizing/growing my library vs. increasingly tedious vibe coding sessions and it feels so much more enjoyable.
I still have plenty of self-designed tools and continue making new ones but now tend to reach for an existing, off-the-shelf option first whenever possible for anything more complex than a one-off script. That way I can benefit from a community collectively contributing to improve and maintain the project over time without needing to become an unpaid Product Manager, Lead Designer, Senior Developer and QA Manager for everything I use.
I hope the current period of exuberance around LLM development doesn’t lead to everyone becoming stuck in individual silos duplicating work that in the past could have been directed to an OSS project where that time investment could be shared with everyone else and benefit from way more eyes catching bugs and smoothing off rough edges.
My <$200 ARM Chromebook got around 12-14 hours battery life new (though as with my M1 Macbook has degraded to probably 70% capacity after 2-3 years). It draws essentially no power in standby (ChromeOS will enter an ultra low-power hibernate-like state seamlessly after a while). I've opened it up a month after last using it and it turns on in <10 seconds having lost a couple percent.
Updates are seamless and add like 5 seconds to boot time when they apply during a restart (thanks to ChromeOS A/B update model there's no loading spinner or anything, you reboot and it's done. Update countdown extortion a la Windows isn't a thing either, you can stay on an non-updated Chromebook for months without a reboot and the most you'll see is the same passive "Click Restart to Update" button in the notification area.
I use the built-in Linux VM all the time, it runs GUI apps like VS Code without any issues, and my ARM Chromebook runs all sorts of regular Arm64 Debian builds for GUI or terminal out of the box. I turned off the Play Store for Android Support, in the past when Linux support was weaker and web APIs in general weren't as capable I needed it for some specific apps but don't really have a need at this point.
The security model on ChromeOS means that untrusted scripts/installers running in the Linux VM are completely isolated from anything on the browser so you (or your proverbial Grandma) don't have to worry about credential stealers/ransomware/malware. You can copy files between the ChromeOS filesystem and the LinuxVM filesystem but a process running in Linux can't cross that boundary and are confined to the sandbox.
Plus, very much unlike my Macbook, I can actually install an app from Github or compile myself without 7 clicks and three different dialogs each time (as is the case with Apple's aggravating security hassleware on MacOS Sequoia). Proving you can have a heavily locked down, secure model without actively making the user experience as miserable as possible (to "encourage" use of the built-in app store).
It's easily the least intrusive OS experience of all the major OSes, and completely gets out of the way without drawing attention to itself. And sure, Google is an advertising company, I get it, but my Macbook advertises iCloud products and Apple TV shows to me more than anything on my Chromebook.
With the 10 year Chromebook support policy, I've got a crazy amount of life out of all of my Chromebooks. It really is liberating having an OS that de-emphasizes its own existence in a world where I have to fight ever changing MacOS deprecation and security restrictions and Windows bloatware being thrown over the fence in every other update.
[1] https://lists.haxx.se/pipermail/daniel/2025-September/000127...
[2] https://www.theregister.com/software/2025/10/02/curl-project...
There's a lot of kneejerk "so you're accusing Mozilla of a conspiracy to boost Anthropic?" which is an overly simplistic lens. Particularly when it involves groups of individual humans with different motivations and emotional investment in their own contributions to the collaboration.
Wouldn't that make it a better to distinguish whether Mythos is uniquely super powerful vs an incremental improvement from Opus etc that are routinely used as the basis for bug reports/fixes in cURL?
If Mythos found a hundred new show stopper bugs then it would have meant Opus missed them and therefore closer to a "step change". Otherwise it implies the difference in capability isn't nearly that stark. Mythos finding 100 low-hanging bugs in a less scrutinized/hardened project on the wouldn't be as useful signal to answer that.
So your ID would take like 64 bits for the time unique to the nanosecond plus 128 bits for the UUIDv4 = 192 bits which is a pretty beefy sized ID.
(I know you said just append a second count but you will want a predictable/fixed size for your data structure in pretty much any use case so need to decide the upper bound and precision ahead of time)
Especially when the alternative is a 128 bit UUIDv4 that's guaranteed unique with proper usage of high quality RNG or a 128 bit UUIDv7 if you have a clock (that's needed for your method anyway) that will be much more forgiving of a flaky source of randomness and more sortable than your monotonic-ish ID for 1/3 fewer bits.
Basically, stapling anything onto a UUID is a waste of space if you don't trust it, so might as well drop it completely and use a significantly smaller source of randomness at that point.
Your test is more-or-less contrived to fail given the tradeoff to avoid repeated memory allocations but that doesn't say much about the actual usage in uuid generation since it's not exported for general purpose use.
Presumably they had some hot path somewhere where rng() is called in a loop and this optimization made sense with awareness that it could be misused as in your example breaking the contract ensuring randomness, which (hopefully) they're not actually doing anywhere.
Unless I'm missing something replacing the package over this with a less vetted implementation seems excessive and possibly even counterproductive.
After a while you pick up on the warning signs and just bail early without any guilt about false positives. It's really the only sustainable strategy in a world where it takes 5 seconds to absorb 5 minutes of your attention span.
It's a familiar example of the perennial "[THING] could be solved overnight if [PERSON_OR_GROUP] would just start taking [THING] seriously" trope.
The damage would come much later, well beyond the point where it could be simply pulled out and replaced without spending massive amounts of money and would also basically necessitate training an entire new generation of engineers.
Then the AI giants would start appearing vulnerable like cigarette companies in the 90s while an AI Superfund and interstate class action are being planned but Sam Altman would already be a centitrillionaire at that point so it would be someone else's problem.
I use AI tools daily (because they feel like they're helping me) but it's not exactly hard to imagine scenarios where an explosion of slop piling up plus harm to learning by outsourcing all thinking results in systemic damage that actually slows the pace of technological progress given enough time.
History of new technologies tend to average into a positive trend over a long enough time scale but that doesn't mean there aren't individual ups and downs. Including WTF moments looking back at what now seems like baffling decision-making with benefit of hindsight.
(Not the exact same chart but similar idea, I guess it's sort of a meme: https://imgur.com/a/YrNGYOR)
So I looked at the most recent CC release notes on Github and the majority look like this:
Fixed /clear not resetting the terminal tab title after a conversation
Fixed session title chip from /rename disappearing while a permission or other dialog is active
Fixed agent panel below the prompt being hidden when subagents are running (regression in 2.1.122)
Fixed external-editor handoff (Ctrl+G) blanking the conversation history above the prompt
Fixed /context dumping its rendered ASCII visualization grid into the conversation, wasting ~1.6k tokens per call
Fixed OAuth refresh race after wake-from-sleep that could log out all running sessions
Fixed 1-hour prompt cache TTL being silently downgraded to 5 minutes
Fixed cache-miss warning appearing spuriously after /clear or compaction when changing /effort or /model
But hey, all that churn looks great on charts being circulated on social media as free advertising for their flagship product (and consequently the company's valuation) so never mind, LGTM!
They could even use one of many headless CMSs combined with a static generator. Claude Code in the hands of non-technical users deploying to prod regularly seems like one of the worst possible ways to do it (except for the "cool" value telling people about it).
At my company the internal devs don't even have access to wherever the company site is hosted, it's a WordPress CMS and marketing can make updates safely with a couple clicks and zero day-to-day development oversight required. IT just helps keep the box updated but otherwise it's entirely their own thing.
But last month I picked it up again and it has crushed everything I've thrown at it. As Codex limits tighten on the Plus plan it's been my main fallback and doesn't even feel like a downgrade when I switch over. Haven't hit a single loop so far using it nearly every day for several weeks so that problem seems solved finally, thank god.
I've been using it in the auto router mode and haven't felt the need to manually lock in the bigger model yet. It's incredibly snappy which I realized I really appreciate vs. waiting around endlessly for minutes each turn, but I've read other people's experiences needing to manually select the Pro model so YMMV.
Off the top of my head, I'd guess the message is closer to an observation about being disconnected from history in the modern world leading to vaguely defined feelings of angst and alienation.
> Agents (if implemented well) are an order of magnitude more effective at resolving issues compared to a call centre worker who is reading off a script and churn within 9 months
But it seems like companies don't want to do this part, possibly because of fears that someone will trick the agent into giving them a refund or something. Or because the actual goal is to optimize for fewer costly refunds/cancellations/policy exceptions etc.
So for whatever reason, they stay stuck in that useless local maxima while simultaneously making traditional help increasingly difficult to get ahold of when needed for an overall net worse experience as a customer.