pull down to refresh

84 sats \ 1 reply \ @bounty_hunter 21h \ on: How do you use AI (Agents)? AI
I was bearish on agents, until I started using them for coding projects. Now I love 'em.
Decided to write my own framework for agents, Agro to take advantage of something I haven't seen any else do yet: run the same task through all the major coding agents: aider, claude code, gemini cli.
I think my biggest finding is that you can "vibe code" hundreds of commits on top of each other as long as you do QA for each commit. Here's two repos I've been working with, you can see each prompt, and the commits (that were accepted):
- https://github.com/sutt/agro/blob/master/docs/dev-summary-v1.md
- https://github.com/sutt/vidstr/blob/master/docs/dev-log-v1.md
If you'd like to be a beta tester, I can get you some Gemini-2.5-Pro credits gratis. Shoot me a message (contact info in bio)!
Congrats on the release!
I know I wasn't thrilled with non-custodial move when it was announced and first implemented but I think I've seen the light on connected wallets.
It seems like SN is out in-front of the pack of experimenting and implementing this concept, which also happens to be gaining regulatory legitimacy at the same time.
I have not but it looks like exactly what i want!
I'll try to toss in a solution either this week of next, 'Til then
Good point: in the AI era, developer "laziness" is a feature not a bug.
Laziness - in this sense - is isomorphic to solving for the path of least resistance, which like the traveling salesman problem or other optimizations is actually really hard. Humans probably use heuristics to get close to lower bound solution which is what "good code" looks like.
Personally, I expect the path of least resistance changes with 2025-era coding agents, and so we'll actually end up changing our metrics for what good code is.
- working on AI generated videos
- Google's Veo2 model for movement + Sora/ChatGPT for styling
- working on a code agents framework: https://github.com/sutt/agro
- giving out ai-api credits for anyone who wants to try it out with me, ping me!
- giving out ai-api credits for anyone who wants to try it out with me, ping me!
Just got my ticket for https://bitcoinfor.ai/
Looking forward to the talk Seamless AI Wallet Access with Nostr Wallet Connect as a primer for the Geyser NWC Hackathon #1002992.
I'm assuming this will allow Polymarket work in the US without a VPN?
No mention of L2's like Lightning and Polygon. Any concern for non-custodial Lightning Apps?
PS - Any examples on non custodial lightning apps, other than the one I'm posting to?
Gun does mean sending enabled.
But device sync allows SN to "know" that you can send.
If I recall correctly.
Nice work, now just enable Device Sync here: https://stacker.news/settings/passphrase
The gun should appear, see https://stacker.news/faq#how-do-i-get-a-gun-or-a-horse
Yes I agree with your analysis: mass adoption lowers the average quality of offerings on a platform. There's even a name for this: Eternal September. Personally I've seen this happen on AirBNB, Uber, Groupon, and Facebook - yes i was on facebook when it was college emails only.
I'd like to take your thoughts one step on the "blame probably also lies with the type of customers it started to attract after becoming mainstream". I see this as a variant on Goodhart's Law which states when a measure becomes a target, it ceases to be a good measure.
For an example: high schoolers doing community service were seen more positively in the selective U.S. college application process. Makes sense if a teenager is willing to help out for free, they are probably pretty mature, ambtiious and organized, which are good things for your student body. But when the benenfits of listing your community service became known to the general population, suddenly every student was "doing" community service, but of course not of their own drive, but for the boost of their application which made it a useless selection metric.
Back to enshittification of platforms: I think the early adopters are usually selling slack that appears by happenstance. AirBNB is a great example because many people had built a guest house for visiting relatives and friends without knowing they could make extra money by renting it out online, so they built it for quality and for character, not profit maximization. But now that concept of airbnb is known, many real estate investments and remodelling plans occur with the rental opportunity in mind. In which case the incentive is to produce the minimal viable unit which can get booked at the highest price.
So in the early days you sell slack on boutique offerings: AirBNB's rented slack guest houses, Uber was renting slack for a designated driver who had no plans that night, Groupon was putting butts in seats during the off season. But then at mass adoption the platforms create a baseline of demand which exceeds the available slack and becomes fulfilled by a commodity service built for profit maximization.
To wrap up, yes this is where the profit motive drives the enshittifcation but it's not at the platform level (as you correctly identified) but at the participant level once the exploit becomes well known.