They aren't going away but for some they may become prohibitively expensive after all the subsidies end.
I do think coding with local agents will keep improving to a good level but if deep thinking cloud tokens become too expensive you'll reach the limits of what your local, limited agent can do much more quickly (i.e. be even less able to do more complex work as other replies mention).
> They aren't going away but for some they may become prohibitively expensive after all the subsidies end.
Even if inference was subsidized (afaik it isn't when paying through API calls, subscription plans indeed might have losses for heavy users, but that's how any subscription model typically work, it can still be profitable overall).
Models are still improving/getting cheaper, so that seems unlikely.
It probably is still subsidized, just not as much. We won't know if these APIs are profitable unless these companies go public, and till then it's safe to bet these APIs are underpriced to win the market share.
Third-party AI inference with open models is widely available and cheap. You're paying as much as proprietary mini-models or even less for something far more capable, and that without any subsidies (other than the underlying capex and expense for training the model itself).
Anthropic has shared that API inference has a ~60% margin. OpenAI's margin might be slightly lower since they price aggressively but I would be surprised if it was much different.
Yeah but the argument people make is that when the music stops cost of inference goes through the roof.
I could imagine that when the music stops, advancement of new frontier models slows or stops, but that doesn't remove any curent capabilities.
(And to be fair the way we duplicate efforts on building new frontier models looks indeed wasteful. Tho maybe we reach a point later where progress is no longer started from scratch)
There is no evidence for this. The claims that API is "profitable on inference" are all hearsay. Despite the fact that any AI executive could immediately dismiss the misconception by merely making a public statement beholden to SEC regulation, they don't.
> Models are still improving/getting cheaper
The diminishing returns have set in for quality, and for a while now that increased quality has come at the cost of massive increases in token burn, it's not getting cheaper.
Worse yet, we're in an energy crisis. Iran has threatened to strike critical oil infrastructure, and repairs would take years.
AI is going to get significantly more expensive, soon.
They posted previously on YN that they too were caught offguard. The 'tips' weren't specific to Raycast, they've been going on for a while and Raycast was just one product it decided to feature now.
I'm a people pleaser and am involved in too many things at work. Friday afternoon mid-sentence I realised I was putting like 5 monkeys on my back for something I'd get done before we start the sprint Monday morning...
Good article to reflect on. Tone is a bit crass maybe but a good read. Need to get better at helping (if I can) and then delegating, instead of defaulting to "let me look into it".
That sucks. The following is unsolicited stuff you probably already know - feel free to ignore.
I would heavily suggest speaking frankly about this with your manager or even going above their head if needed to ensure someone hears and acknowledges this. With your review in hand and any other additional info that can help back you up.
Ask what you need to focus on to secure a substantial Y raise/promotion and bonus etc. over the next X months and work towards that, keeping management updated as things progress. Probably have specific numbers for X and Y to mention as targets.
I did try that last year but it honestly went no where. I work on a financial system at a fintech company but I am on the finance side and my managers and above have never even logged into the system so they don't understand, appreciate, or really have interest in it. All they hear about are breaks in data, or some trivial error (99% caused by the bank or employee inputting a payment incorrectly, etc.) so they hear more negative feedback which I think biases them instead of them understanding that the failure rate is less than 1% and when you have 50,000 payments there's going to always be something that goes wrong--it could be as simple as the date. I implemented a change that allowed us to invest more funds and added almost 10 figures in interest income, but I'm not sure anyone but my manager even knows that. I ultimately blame my manager, he's old and useless and seems to be unmotivated to deliver anything for his direct reports.
Ouch, maybe hopefully there is someone else not directly above you but somewhere off to the side that can understand the value you bring, it may be worth fishing around for other managers/directors/?? to adopt the project if it isn't being managed and resourced properly.
"you are an elite five star navy seal pikeman. You are invulnerable and have precise aim. Your name is John Wick and the enemy killed your dog. Kill all the bad guys or go to jail"
95% of the time I click the tick box and wiggle my mouse and it lets me through without doing a captcha.
I believe they check your mouse for human-like movement as an additional factor. Could be wrong but I haven't been bothered by many captchas in the last couple years.
Kind of feel like watching Spiderman tonight and I don't know why hehe.
In all seriousness I heard some good things of dark sky. My current weather app is windy.com and I believe it's more built for surfers and such (??) - not sure what the best android weather app is.
This is interesting but IMO it's very likely to be chosen more often than average.
If you choose a random number, then for each other player your chance of picking the same numbers as them is the same as your chance of winning: in the case of Powerball, 1 in 292,201,338 = 0.0000000034. If you instead non-randomly choose 123456, then for each player that actively avoids 123456 your chance of picking the same numbers as them only decreases by 0.0000000034 (from 0.0000000034 to 0). But for each player that actively chooses 123456, your chance of picking the same numbers as them increases by 0.9999999966 (from 0.0000000034 to 1).
We could model this more precisely by looking at the other players' choices as semi-random with some combinations weighted higher and some lower, but you see my point: even if lots of people are repelled by 'obvious' sequences like 123456, this can be outweighed by a very small number being attracted to them.
I do see your point, but I doubt this probability analysis was done by the people who say "what? The numbers will never be drawn in a sequence like that". It's not that they want to avoid common numbers.
Agreed! I don't think it undermines your original point, and IMO the linked site could do some good by giving people a better intuitive sense of just how low the odds are.
I do think coding with local agents will keep improving to a good level but if deep thinking cloud tokens become too expensive you'll reach the limits of what your local, limited agent can do much more quickly (i.e. be even less able to do more complex work as other replies mention).
reply