Not sure why author didn't use 4o-mini. 4o for reasoning but things like parsing/summarizing can be done by cheaper models with little loss in quality.
I have apps that pay my rent such as
YOU-TLDR - Transcribe and Summarize YouTube videos
you-tldr.com
Shorts Generator - Text to Automated Shorts In Minutes
shortsgenerator.com
Snoop Hawk - Automated Web adn Reddit Marketing
snoophawk.com
You can find all my projects here
hackyexperiments.com
Currently I am exploring the idea of chatting with more than one AI. My assumption is it makes the interaction more life like and less lonely and I think in the future everyone will hav a group of AI friends. I made a quick loom about it here: https://x.com/deepwhitman/status/1826831221554643324
I love this pattern as well, but as some comments have pointed out, there is a pitfall of security risk (you don't know where your API key is going). I think it would be really cool if more services let you cap API key usage, so it's almost like a virtual credit card number. I could then put in API keys to sites without worrying and knowing that at most, I can lose, say, $10 or something.
This is absolutely the thing I most want: it should be trivially easy to create a dedicated API key with e.g. a $5 maximum spend that I can share with a "bring your own key" tool.
Being able to issue those via an OAuth flow as opposed to copy-and-paste would be nice, but the fundamental thing I want is per-app spending limits.
Obviously you aware, but let me still say it: openrouter is for me a trivially easy way to create dedicated keys with max limit https://openrouter.ai/credits or am I missing something?
This is pretty much exactly what an oauth2 flow is. You generate a single-use-specific token and constrain it with the limited permissions you choose. If a site wants to offer an oauth token flow where users can put a time or credit usage limit on the token they absolutely can do so.
I have not been able to get it to output anywhere close to the max though (even setting max tokens high). Are there any hacks to use to coax the model to produce longer outputs?
My controversial take on this is that we are in the mid-curve of this tech, i.e., smartphones/social media are not quite there to replace IRL experiences and are even further off from real community...BUT instead of going back, we need to move forward to the right side of the curve where full VR / network states can solve a lot of these problems.
I'm very bullish on IRL experiences. Community building is more complex, with various ripple effects to consider, but realistically we are heading in that direction whether we like it or not. I find it more compelling to explore how we can reclaim and enhance these lost aspects in our modern world rather than going on "back in my day" nostalgia trips.
"I predict one day Amazon will fail. Amazon will go bankrupt," Jeff Bezos.
One of the most surprising things in my brief experience at Amazon was how much of a shitshow it was. The two-pizza rule and self-sustainability for each team led to huge overlaps, with teams doing the same thing.
With such a huge organization, you had to go through 15 different stakeholders to get a single thing done, and there was an ingrained middle management whose only function was to connect you to the right person.
Just figuring out who was responsible for what and how to get things done was a challenge in itself.
Despite all this, Amazon still succeeds, and their process of PR/FAQ, leadership principles, and one-pagers is one of the best I've ever seen.
But I wonder if at some point, like with any philosophy and just like Bezos predicted, it will become too much and the whole thing will cave in on itself.
To give Amazon and Bezos credit, their 14 leadership principles give a pretty balanced framework for decision making. Customer obsession and working backwards to help focus on what's important, disagree and commit to resolve differences that root in intuition or assumptions, bias for action to minimize analysis paralysis, and etc. It's amazing that Amazon can sustain its size for so long. But yeah, eventually it is people who enforces culture, and eventually an empire falls.
The LPs are bullshit. They are basically templates that the people in power can (ab)use to force their narrative. Want others to shut up and do the job? "Disagree and Commit". Don't want to spend money on employee's well-being? "Frugality". All else? "Customer Obsession". And what does "Are Right, A Lot" even mean?
They subject us to meet with them as customers on Chime. This is despite everyone hating the product. CodeCommit was annoying because nobody uses it, and you had to touch adjacent projects like CodeStar to do something like trigger on a GitHub commit.
There's nothing in Yegge's rant that implies Amazon must dog food AWS. AWS as we know it, was, if anything, a byproduct of the efforts described by the rant.
Amazon certainly uses AWS but I think it's foolish to think they use every product (they obviously do not) nor that they must. For the foundational pieces -- EC2, SNS, S3 -- sure, of course. They spent a long time migrating workloads.
I don't expect that with every new service they launch that they have already widely adopted it by then. Maybe after their customers widely adopt it, if ever.
> There's nothing in Yegge's rant that implies Amazon must dog food AWS
Really? From the link:
> "The Golden Rule of platforms is that you Eat Your Own Dogfood."
and from Yegge's interpretation of Bezos' mandate:
> "All [mandated internal APIs to be used for all interservice communication], without exception, must be designed from the ground up to be externalizable. That is to say, the team must plan and design to be able to expose the interface to developers in the outside world. No exceptions."
Of course Amazon doesn't necessarily use every single AWS feature internally; I don't think dogfooding implies building all your features for internal customers first, and then selling them. Rather, it implies refraining from selling crappy reimplementations/duplications of internal tools.
> With such a huge organization, you had to go through 15 different stakeholders to get a single thing done, and there was an ingrained middle management whose only function was to connect you to the right person.
If you're able to recall, can you list who these 15 stakeholders were? Are they like legal department approval, etc?
reply