296 karma · joined August 2, 2022
for me just having a brushless motor and a remote was good enough lol. maybe smart home functionality nowadays perhaps.
also, all of this seems like crying wolf as usual, which won't help for when we finally get there. lowkey, following this noise is getting as horrible as getting all your news from linkedin posts.
meanwhile the downgrade to the 200 plan, while not as outrageous as github copilot, is just not good pr fwiw. the base number, i.e. what plus users get, is arbitrary as it is, so there was no need to break the appearance of "20x". using the efficiency excuse is not helping here lol. same as further segmentation of pro plans with ultrafast mode.
plug-in hybrid don't make sense to me based on the op's experience, but the regular hybrids (where you don't charge the battery) makes perfect sense today, more than ever arguably. same tech as f1 cars fwiw (not that this season they are being very favoured)
while not my cup of tea and not going to rewatch it, i see a repeat of a lot of the ways people react to "world-changing" topics.
or in other words, a lot of hand-waving since past few years is only going to be crying wolf when maybe we ever get there. perhaps before nuclear fusion or gta 6!
for all the promise, i would temper my expectations for a frontier model for coding or general purpose. but for niche use cases (document parsing, voice*, etc.) and as a platform to handle finetuned versions, etc. for small-ish pilots, i can see the value of partnering with them.
if the core model could be any open-weights solution out there, the moat dissolves really quickly.
Sadly, things are not quite as straightforward for non-Studio. For one, you need to be a paying customer to get access to the new feature. Moreover, a community-based solution [1], which previously worked for all users, have now artificial limitation in implementing automation for Free users.
Quite a shame really but maybe it will trickle down over time (like expanding export resolution support). Regardless, a much more open implementation than Adobe, who'd like you to pay for yet another chat window.
[1] https://github.com/samuelgursky/davinci-resolve-mcp#free-edi...
i wonder when the migration was started and whether it made sense to not migrate it to another framework instead.
however, a lot of creators in it had direct ownership. i wonder how it works for those who joined later on and may not share the same.
one of the creators used to be super-active, however now switched the style of videos (completely fine and up to the creator), but now it has gone to radio silence for a better part of a year.
maybe the current creator economy is so bad that once you can get a gig like nebula, it pays you enough to stop the treadmill of churning out content.
to be able to maximize customized solutions, they would like to own the entire stack which is not already commodified. the cpu remains the only final frontier in this.
there are always side-effects possible, but i wonder if all these efforts can be applied outside of ai ever.
the new price is welcome but still 60% more expensive than 5.4 (which is now comparable to 5.6-terra).
i conveniently also leave out the "-pro" tiers, which was comically expensive. but i feel like openai consider them suitably comparable to sol, and it must be them offering a huge bargain.
outside of coding scenarios, anthropic is just not quite the best choice for regular folks.
depending on the cloud/office stack used by the company, the userbase is better poised to use the native ai offering than moving to anthropic and connecting everything with them. same goes for third party guys trying to tie it all together.
not to discount them or their features, but beyond small companies or startups, we don't see the same levels of adoptions now nor maybe in the near future.
those asking us to build genai solutions for them also do not choose anthropic first over the main model offered by their current CSP. anthropic made itself available across the board finally, but it was already too late for most projects planned out for the upcoming quarters. also, we don't use the absolute biggest model nor instantly migrate to the latest one. the insane price jumps do not help with the same. especially when we quote something that may no longer be true down the line.
aside: good luck to the FDE folks hired by the frontier labs. i wonder if they would really get to handle large company-wide projects in practice.
the programmer, especially with crocodile clips, can be used for other laptops with bios chips of the size. however, it is always nice to have info from someone else who has previously done this. while it took a lot of digging for the specific laptop i had, at least with framework we hope to see more high-quality writeup like this.
ever since i had my own money, gpu or general hardware had some excuse to fleece us. however, with ram the blatant collusion throughout history does not invite confidence.
this looks as dramatic as it does because the recent trend happened right after a major crash in prices. throughout the years, the usual price for an 8gb stick would hover around 50-80 usd (albeit laptop one, might be lower for desktop). ddr5 also started being 50-100% more expensive than the incumbent when it first dropped. it took a better part of a year to be in a similar price bracket as ddr4.
all this to say, this situation sucks but those of us with 64+ gigs, bought during the last crash, are eating good but got used to using much more memory than they really need.
while it is insane to pay a decent cpu price for 16 gigs today, getting a "good enough" computer still makes more economic sense than a max plan from a frontier lab.
i don't think non-fb apps are only running the meta's ad revenue, which is meant to overtake google's.
while i personally don't use it, does not mean they are not getting enough eyeballs.
however, found it interesting that fb was able to quietly win the battle. however, it is a shame that nothing can be done on the mobile front, at least inside the apps. it is insane how sometimes every other thing you see is a sponsored post or an ad.
we are blessed to be in an arbitrary field where we are free to steal ideas from other disciplines and find new innovations in our own.
to use it, open the meatballs menu (...) of your current copilot conversation and click "show agent debug logs". it opens a tab showing all the various tool calls and prompts being sent out behind the scenes and how model selection is happening (if auto). it also gives insights on token consumption as well. moreover, vscode has been quietly shipping updates and recently you can connect your own otel service to get this information in a way you can put to use yourself.
while copilot has turned me off post their pricing changes, they have been doing tons at their own pace. highly recommend going through this feature if you find the op interesting.
time for another approach to run your agent's web searches through this (or by mocking browser signature), with potential cf bypass built-in!
simultaneously, i find it curious that the ide's git integration feels incomplete in front of vs code and its forks.
git is not perfect but for a lot of users, a better ux can help solve most of their grievances. regardless, network effect exists and it would take more than a killer app/site to get this going.
i would not outright dismiss this attempt, just like svn guys may have treated git back in the day.
similar to cpu benchmarks (e.g. geekbench, which also had a recent update), the goal is to distinguish model in some arbitrary thing. however, it is not quantifiable but subjectively even the latest show distinctive differences regardless.
large context window is still relatively new and not "feasible" even commercially, so making that a requirement for a benchmark would not make it accessible especially for open-source ones.
* yes, i misspelled it intentionally
case in point 5.4-nano, which is the only model of its class which is now finally comparable to 5.6-luna in pricing. [1] until this change, workflows using the older model could not economically justify the old luna pricing.
the doubling in price between 5.4 and 5.5 also did not help with things, but now 5.6-terra can be considered an update for this pricing tier.
if the token efficiency is real (which i am yet to replicate for our workflows), then the switch to these models can be a net benefit. thank you chinese labs!
[1] https://developers.openai.com/api/docs/pricing#text-tokens
the benchmarks right below the list of models is weirdly set. very unlike the standards of the usual alphabet websites.
have seen this with copilot, then with openai, and usually they tend to be bombarded with some examples at first. only to be not maintained down the line or become irrelevant.
however, lately the anti-bot/spam shenanigans from big tech has ruined any ux work they have spent trillions on.
<rant>
one fine day, i was trying to log into my google account at google's office. for some reason their internal wifi was serving the internet through a corporate vpn, which was based out of the US. this tripped their security system, which then made me jump 3-4 hoops to log in back to my account. i recall sms verification (where you send them a code), some part where i had to scan a qr code and open on my phone (not logged into google btw), and the usual 2fa, among others.
likewise ticketmaster asking for verification for a decently used account over the years out of the blue. this is when booking a new ticket from a residential internet without any shady business. was made to reset my password three times in a row, and still got blocked!!!
</rant>
the false positive rate of these systems have reached a boiling point at some instances. yet they don't seem to stop the spammers and scalpers, but cause great pain to regular folks who don't pass the sieve. passkey-based verification is just a symptom of the larger problem.
PS: don't even bother using a non-normie browser, as it is a sure-shot way to get restricted, even on this website. it is as if we are all using tor browser if it is not the latest chrome.