Currently in stealth mode
Imagine you have no token limit. How would you orchestrate CC?
With all the hype of fable 5 I have the FOMO that I’m not utilising my tooling fully.
What I do:
- many custom skills
- memory management
- fan outs
- some parallel sessions
But I never burn absurd amounts of tokens and neither are my sessions running more than 15 minutes on average. I tried opus as well as the higher reasoning efforts. I found it a waste of time and tokens.
Maybe my scoping of tasks is too narrow? Whenever I tried giving it bigger and more ambiguous tasks it was lazy or took wrong directions.
Claude Code gets 403d constantly when fetching URLs like docs, Reddit etc.
I know you can solve this with playwright but I was too lazy to set it up and just used ChatGPT / Gemini for (Deep) Research.
But then I got too lazy to switch apps. So I thought lets configure it properly such that Claude Code can use my browser session for ChatGPT/Gemini.
Quick walkthrough incl. claude skill:
One thing people don’t talk about is how great it can be to finally stop a project that wasn’t getting traction. Even greater when you have a promising idea for what comes next :)
Anyone with experience in finding and working with design partners at the very early stage of a project? Any learnings or tips?
Are you using your unfair advantage?
Most "startup idea databases" are useless because they ignore Founder-Market Fit. You get 500 people building the exact same generic SaaS.
What if a system parsed your CV/GitHub etc., found your asymmetric domain edge, and matched it to live, niche internet complaint clusters?
If a tool showed you a massive B2B problem pattern that only your specific background could solve, would you build it? Why wouldn’t you?
Let me know what you think.
Are you building B2B or B2C?
Feels like most here are in the B2C game. Curious whether that’s true and why you have chosen the one over to other?
Personally, after some B2C projects, I think I’m going to try B2B next.
I fell into the trap of building something no one asked for. So I decided to do something radical:
Drop everything that isn’t solving the core problem.
The app is now so borderline simple that I’m ashamed. At least I improved the speech2speech latency by a factor of 3.
Check it out, feedback welcome: apps.apple.com/de/app/ai-mental-health-withease…