
@simonbalfei
iAccount based inUnited Kingdom
About this account
- Account based in
- United Kingdom
- Connected via
- United Kingdom Android App
Account-level information from X, not a live location or the device used for a specific post.
social media scraper for ai agents - https://nitter.cf/t.co/exK2v8utVv
Joined July 2024
- Tweets3.6K
- Following389
- Followers479
- Likes6.5K
For anyone who uses AI a lot to code.
Install "ponytail".
It gives you suggestions on how to reduce the lines of code and dependencies.
This is great at "de sloppifying" your code.
recently had some crazy health condition that makes it dizzy and confusing to walk outside sometimes.
its insane how much you take for granted a basic thing like this
but you know, stephen hawking developed hawking radiation while being crippled
so this is nothing
to be honest the biggest lesson in indie building stuff is just copy something making money then spend 90% of the time mastering distribution for it.
learning ads, seo, cold outreach is 90% of the complexity and this is a field ai vibe slopping doesnt fully solve
this means dont make something super complex really
if the market is big enough for the problem being solved then theres a 99% you make some money
just keep repeating those till you find one that clicks.
passed $1000 made with my saas CreatorCrawl.
not bad for the first proper project.
we created a stupid new protocol and now we are complaining it has problems lol.
modern engineering.
mcp is the dumbest default for gtm workflows.
here is the usual mcp flow:
• pull a company list
• ask claude to filter for your icp
• find people at the survivors
• ask claude to filter for the right roles
• enrich the keepers with emails
but if your icp is “uk saas companies with 50+ employees”, claude is not making a judgement call.
it is reading rows and retyping a spreadsheet between tools.
the cli version is one pipeline:
• get companies
• filter the rows
• find people
• filter the roles
• enrich the survivors
claude writes it once. the shell moves the data from then on.
no repeated ai reads.
no retyping.
no token cost between every step.
use ai for actual judgement. use cli for deterministic filtering and data movement.
claude should be the plumber, not the pipe.
i feel like it can look like a market is saturated (good anyways right, demand)
but if you actually dig into the massive array of tools, most are probably sub par and easy to beat.
especially in the vibe coding era, where getting a scrappy product done fast is more valued than a product with legit depth and value to it.
commercial cleaning cold email is kinda cracked.
sent like 400 emails and got 8 positive replies already.
boring ahh idea but gets paid.
It's very easy to charge less than you deserve because you have a skill that is trivial for you to do, but for someone else it's a complete game changer.
cli > mcp
so much faster and more efficient
just made a cli for google ads and its secured conversions already at a cheap price
Released my SaaS (CreatorCrawl) a couple of months ago.
Setup SEO and email marketing.
Left it for a bit to focus on some other agency stuff.
Secured $500 passively without much marketing.
Someone copped the $99 deal even :D.
Pretty cool, If I focused on it more It could make a fair bit more.
Happy so far for my first proper project though.







