it has become this due to the cyclical mania and bad margins on the troughs killing companies. hopefully this current hunger for memory make it less painful down the line.
298 karma · joined August 2, 2022
it has become this due to the cyclical mania and bad margins on the troughs killing companies. hopefully this current hunger for memory make it less painful down the line.
what made you migrate from Drupal?
from past videos i could recall your vast experience with running websites using it, even using drupal deployment for some sysadmin tutorials.
much appreciate the insider baseball about your work!
EDIT: there is a post on this - https://news.ycombinator.com/item?id=46487498
in the past, cropping out a part of a video would meant reencoding it in some random preset. this would often take longer than required. however, accidentally realized the difference when trying out avidemux [1] and clipping together videos blazing fast (provided in same container and format)!
as pointed out by others and throughout the thread, anti-cheat is very restrictive inside a vm. even cloud streaming fails to support some popular titles from EA and R* due to this.
meanwhile, WSL exists, and now provides good gpu passthrough and has a higher success rate with the said permutation.
while LLM-generated text was already a thing of the past couple years, this year images and videos had the "AI or not" moment. it appears to have a bigger impact than our myopic world of software. another trend towards the end of the year was around "vibe training" of new (albeit much smaller) AI models.
personally, getting up and running with a project has been easier than ever, but unlike OP, i don't share the same excitement to make anymore. perhaps vibe coding with a phone will get more streamlined with a killer app in 2026.
the time requirement alone makes it tricky to be used directly at the source of carbon emissions. this might be efficient but unsure about effective.
- the bugs feed on crop waste instead of the animal waste of the target animals [1]
- they are also investing in vertical farming alongside the core business
still not profitable as of writing the linked article, however.
while i don't agree with how devs and the publisher works on community feedback, it is still miles better than what EA does. not that it is a high bar to clear.
conda already had the independence from python distribution, but it still had its own set of problems with overlap with pip (see mamba).
i personally use uv for projects at work, but for smaller projects, `requirements.txt` feel more readable than the `toml` and `uv.lock`. in the spirit of encouraging best practices, it is still probably simpler to do it with older tools. but larger projects definitely benefit, such as in building container images.
mine are rehash from this year (which failed!) with a new one:
- focus on health: specifically getting back on track and maintaining a sustainable routine
- consume less, make more: less passive pastime
- explore side hustles: to encourage the previous, try to fill in a (very) small market gap using what i know or want to learn
> Turns out Claude Opus 4.5 knows the trick where you can add .patch to any commit on GitHub to get the author’s unredacted email address (I’ve redacted it above).
given how capable certain aspects of these models are becoming over time, the user's intent is more important than ever. the resulting email content appears like a poorly-made spam (without the phishing parts), while able to contact someone just from their name!
regardless of the semantics, git is not ideal for serving files. this has been more apparent in the ai world, where extensions such as git lfs has allowed larger file size.
but as seen elsewhere, network effects trump over any design issues. we can always introduce an "lfs" for better shallow fetching (cached? compressed?) and this would resolve a majority of the op's grievences.
macos and linux usually come with a python installation out of the box. windows should be following suite but regardless, using uv vs venv is not that different for most users. in fact to use uv in a project, `uv venv` seems like a prerequisite.
i completely understand the archive's decision on applying their own compression.
as a fairphone of laptops, it still seems like a much better deal imho. the core performance is not much worse than the competition. the user-facing hardware, while not as good as the best, it is quite in the same ballpark as other windows or linux laptops.
also a lot of quirks are starting to be solved by the community, which seems to be in a healthy state. the pixels and macbooks have enjoyed this in the past, despite their own quirks mainly due to network effects. plus rarely one could opt out parts not needed while ordering the laptop.
the edge/on-device AI use cases on smartphones can also extend without user friction through web apps built on the above standards. perhaps one day there will be a "WebNPU" or just get supported through existing standards.
there are already some use cases on apps but it usually fallbacks on cpu. perhaps it could be the hw accelerated moment that we saw with video on the web.
personally, as someone building on top of gen AI for a living, i finally bit the bullet on building using LLMs. it did reduce friction in things i don't like doing and did not explore as much. by acting as a catalyst when i needed to finally address them, it helped me get going and eventually become proficient in the core tech itself.
outside of work, however, i find people around me use the services much more than i do. sometimes it felt like the "big data is like teenage sex"[1], but some aspects were quite genuine. got better appreciation after trying them to better understand other people's perspective and to design better.
with "slop" as word of the year and people wondering if a random clip is AI, now more than ever the effects in general life seems apparent. it is not as sexy as "i will lose my job soon", but the effects are here and now. while the next year will be even more interesting, i can't wait for the bubble to burst.
most of the discourse here is missing the point. it is not to migrate a cloud resource away from (assumed) US-based services, but rather moving to the cloud for core services.
now "sovereign" is the hot word in these parts, and of course it would be a priority. but if it ends up running American HW/SW, is it really being independent? regardless, the article also mentions Google workspace, which is not in the scope of migration.
regardless of which part of the fence one is in, things are easier said than done especially when IT services is a cost to the company rather than revenue stream.
- benchmarks don't mean a lot for the frontier stuff, but can be interesting for the same series of models (smaller v/s larger). reminds me of comparing clock speeds between CPUs.
- the app layer can fill the gaps to squeeze out the most for a use case, but there is still no one-size-fits-all situation.
- often the discourse here or the perspective of people building seem disconnected from an average user. a lot of discussion in the post is irrelevant for the vast majority of users. e.g. as cool as TUI can be, it is not an interface most users would gravitate towards.
while not directly related, other modalities are more exciting, and comes thanks to applying techniques for handling text to other media forms, or in conjunction.
about privacy concerns - if you limit it to your work (and if your company is cool with data leakage risks), you can still do things like the video shows.
i do wonder if there could be more potential use cases if the underlying models also support audio. not for user input but rather audio playing in the browser.
> /audio-features/{id} "Get audio feature information for a single track identified by its unique Spotify ID."
this combined with track metadata can finally allow those motivated enough to create their own personalized shuffle. potentially better than the slop we get nowadays. no generative ai required*.
while one can compare in terms of number of tracks, the quality used to be in another level altogether. from the article:
> The quality is the original OGG Vorbis at 160kbit/s.
meanwhile the tracker had 16/24-bit flac rips of vinyl, with decent quality control where the track's metadata was verified for any artifacts. for the given quality, one could rip youtube music (maybe not as easily anymore) and achieve a larger scale in a similar quality level.
now if hypothetically tidal had all the music of the world and was accessible this way, then it would be a comparable resource. insane regardless.
simultaneously there are many opportunities throughout to harden one's app to avoid similar exploits.
call me the worst junior dev in the industry, but pre-coding agents, closing tickets was more important than upholding absolute quality. not everybody is dealing with a billion concurrent users with multi-geo deployments. most of the time, a few screenshots or test output for manual validation is enough to go ahead. when pressed with time and without the prerequisites in the infra side, doing the absolute best development and testing is a luxury only for daydreamers.
automated testing can be a double-edged sword. pre-LLM, even test coverage was a number that somehow needed to go up after each PR. this only resulted in shady tactics of pointless test cases that slowly bring up the metric. today it can be very dangerous if both code and its test suite are vibe coded. especially when it can give the appearance of that 90%+ code coverage.
on the other hand, some manual testing to make sure the core functionality works is the bare minimum one does before pushing out code. at least i would like to believe it is.
i am not even through the entire report but already spotted some dark horses running the web without much fanfare.
it is a shame that it needs to be this way. as a lurker who doesn't stay logged in nor use incognito mode, i have seen "Sorry" page way too often, even when opening the "past" page from the homepage.
truly hope you find a solution that reduces friction for all. personally, it is back to "Sorry" situation for now.
PS: for others facing a similar situation, it all disappears after logging in, which has been the most reliable solution thus far.