Bardia

Dreamer and AI engineer

[Bardia @bardiakha]

Gemini feels like a good kind and experienced teacher. I go there for looking up things like definitions or small educational sessions. The UI feels better for that type of work.
Claude feels like a good smart engineer, feels a bit too dry and serious for when you wanna do simple learnings.

[Bardia @bardiakha]

I’m 34 and this is gold even for me!

[Tim Denning @Tim_Denning]

https://t.co/fnire2eOJr

[Bardia @bardiakha]

@imjustnewatai Damn!!

[Mark Tenenholtz @marktenenholtz]

“Dad, what was it like to start a new project on AWS in 2025?”

[Bardia @bardiakha]

I’m using my agentic software factory to build my agentic software factory.

[Bardia @bardiakha]

Well said!
Your system is as fast as the slowest component. Agent-based pipelines can’t afford being as slow as a human code reviewer.

[DHH @dhh]

It takes a while to build the confidence, but there's no future where you're manually reviewing every line of agent code. Not much acceleration in that. You need adversarial agent reviews, you need automated testing, and maybe you spot check. That's it. From prompt to production!

[Bardia @bardiakha]

@Polymarket Wasn’t he saying the same stuff about Metaverse? Come on man, your device would be the last thing I’ll be intimate with!

[Bardia @bardiakha]

Took the SW industry 5 years to go from:
“Can we build this?”
To
“Should we build this?”

[John Kim @PremiumGoblin]

We need a graph for abandoned side projects since AI coding.

When a new idea is just a prompt away to prototype I find myself chasing multiple ideas that I end up throwing away.

Having diligence to find an idea to work on for the long term is getting really hard.

[Bardia @bardiakha]

A lot of habits die before getting born, took me 3 decades to learn:

1. Routine: do it every day, even for 5 minutes
2. Quantity: increase the volume as days pass and your time allows
3. Quality: refine your techniques and work on quality wh

[Bardia @bardiakha]

@kunchenguid I was watching your first-mate video on YouTube today and the backpass totally got my attention can’t wait to try it

[Bardia @bardiakha]

Glad I was part of the first and last generation of SW Developers.
History will remember us!

[Bardia @bardiakha]

Put down that context from your head into docs before starting with tasks!!

Simple, but magic!

[Bardia @bardiakha]

The only thing between an agent and a destructive or outward git command is the auto-mode's classifier, and it is inconsistent!
My Opus 5.5 today ran the same 'git restore' twice: the first time the classifier blocked it, then I asked for a retry and this time went through!

[Retro Coast @RetroCoast]

Voting is the adult version of writing a letter to Santa Claus

[Bardia @bardiakha]

How silly is it to ask models to compare themselves?
1- Give Opus 5 and Opus 5.5 the same prompt for the same task ( mine was to review a PR and categorize the changes into three levels with reasoning.)
2- Exchange their answers to review each other's output with fairness
3- Ask them to identify the differences
4- Ask them to sort the differences by importance
5- Ask them to grade each other's work

This is what they gave in my case:
Both Opus5 and Opus5.5 gave better scores to Opus5.5!

Opus 5 gave itself a 6 and Opus 5.5 an 8. Opus 5.5 gave itself a 7 and
Opus 5 a 6.5.

It's funny how each model is humbling, scoring itself