Samani

does anyone else's claude code sign them out every single day?? @claudeai please fix your app

AI agents are making it ridiculously easy to build software. but getting customers is still by far the hardest part. so we built an entire AI sales team that researches leads, qualifies them, calls them, books meetings, and follows up for 30 days. all running completely automated. here's what the system looks like in action, and a full breakdown of how we built it below. teapawt.org/samani_/status/2115475…

having AGI in every pocket will NOT even out the playing field the people with skill and initiative will get a 100x multiplier, and everyone else gets a better autocomplete we've seen this before: social media handed every person, in every field, free distribution. a few built audiences, businesses and careers off it. but most just doomscrolled..

this is so important. the right 1-2 friends do more for you than any course, coach or productivity system holding each other accountable means you can't hide behind "i'm working on it" when they're asking what you shipped every single day pick your people carefully

You don't need to spend less, you need to make more

woww. T3 Code is honestly incredible... as much as people love to hate on @theo, he's done an incredible job with this app. so much attention to detail: - switch between models in the same chat without losing context - use your existing claude and chatgpt subscriptions - import / sync chats from other providers - track token usage, model costs and session limits across providers - old sessions automatically get “settled” so your workspace stays tidy - clearly see which sessions belong to each project and which models you used and somehow the UI is still clean. hands down the best-looking harness i've ever used. i literally built a bridge between claude code and codex because switching tools was such a pain. seeing this much thought go into that experience is soo good. and before anyone asks, no, this isn't sponsored lol teapawt.org/samani_/status/2111191…

Just built a local bridge between claude code and codex so when one hits its usage limit, i just switch to the other and it already knows everything that just happened I've got it running on hooks so before every message, each app gets fed what happened in the other one finally don't have to keep re-explaining and copy-pasting the project/session context

if you've been coding with agents for a while, you NEED to try this it just freed up ~200gb on my Mac! 47gb - old agent worktrees 37gb - npm/pnpm cache 24gb - node_modules + builds of old repos 21gb - unused Docker images 58gb - other app caches + junk paste this into Claude Code: "My Mac is running out of storage. Audit what's using the space and give me a ranked list of the biggest, safest wins, with what each item is and whether it rebuilds itself or is real data. Pay special attention to dev clutter: leftover git worktrees, package caches, Docker, node_modules and build output, and leftovers from apps I've uninstalled. Don't delete anything until I approve. When I do, back up any uncommitted work first, and give me anything that needs sudo to run myself." trust me try it now, or bookmark it for later

I think one of the biggest AI opportunities in 2027 is hiding in plain sight. every time I open LinkedIn, I'm seeing more companies hiring for AI implementation and automation roles. these companies already want AI, have the budget, and are literally paying ppl $10,000s to figure out how to use it. why not approach them as a fractional AI engineer instead of a full-time employee? with coding agents doing much of the building, you could easilyy take on 3-5 clients at once. I'm seeing this firsthand at Agenix AI. building the AI systems is becoming the easy part. but getting them to reliably work inside a real business with messy workflows, existing systems and actual employees is where things get super interesting. and that's where the money is. AI is making the technical work cheaper, but businesses still need someone who can turn all that capability into actual results I suspect the people who can do both are going to be in a pretty incredible position over the next few years.

What if the jobs AI is creating today are the same jobs it'll replace tomorrow? We're being told to adapt, retrain, and learn to work with AI. But what happens when AI becomes better at those new skills too?

you're missing the whole point @MrCollison. you can't just think about the state of AI "right now" or even the near future. At the rate AI is advancing, the reasons those small businesses stick with the big tools start to disappear. Right now yeah, they won't replace MS 365 or Salesforce with something a random agent spun up. Too much risk and not enough reliability. Makes sense. But if the AI keeps getting better at this pace, it won't just generate the software. It'll keep maintaining it, hardening it, handling the updates, the security, the compliance checks, the support. Continuously. Better than most vendors do today. At that point "someone is on the hook" can just be the system itself plus whatever insurance or audit layer sits on top. The quality and reliability gap closes or flips. Then the fixed SaaS subscriptions for generic functionality get a lot harder to justify, even for the stuff that runs the business. This is basically what @levelsio has been saying. Network effects and pure legal/regulatory friction will still slow some things down. But for most of the software they're paying for every month, the current stickiness is temporary. It's a capability problem, not a permanent one.

Not all heroes wear capes