Dmitry Lezhnev

i make software.

[Dmitry Lezhnev @dimalezhnev]

In fact I do have @unclebobmartin skill, use it all the time, happy as hell :)

[Dmitry Lezhnev @dimalezhnev]

@betoayesa yeah, in fact I find myself thinking hard while the agent is doings things. I have a few concurrent fagents working and actually feel overwhelmed with following them.

My next strategic goal is to master non-HITL agents flow based on ralph loops and beads for planning.

[Dmitry Lezhnev @dimalezhnev]

@GeoffreyHuntley the problem is so real and painful, I'd love to see the direction the solution is at.

[Dmitry Lezhnev @dimalezhnev]

What an agent needs is small footprint of the code, to allow small context window and better LLM reasoning.

It feels right. It feels potential.

[Dmitry Lezhnev @dimalezhnev]

I've been experimenting with no-framework app setups since many years ago (some back reference: https://t.co/qNBUfTFR4Z)

Eventually was gravitated towards big Laravel ecosystem.
But now the old "light and little-dependencies" ideas are coming back to support AI agents.

[Dmitry Lezhnev @dimalezhnev]

@mattpocockuk that makes perfect sense and hits the exact problem it solves (hovewer it does not include the mention quality back pressure, afk can produce wast amount of crap), considering that "Ralph" seems to be off to crypto adventures.

[Dmitry Lezhnev @dimalezhnev]

why spin up a huge PHP/Laravel setup for a project if agents will exhaust its context just by analyzing existing files. Instead light (micro) setups will drive automatic software making by optimizing for efficient context usage, less moving parts, and easier testing.

[Dmitry Lezhnev @dimalezhnev]

How the business will change if any selling product is copyable in days and for a few hundreds of bucks. WHAT is going to happen with the business models, companies and markets?

I wonder if I can ask @GeoffreyHuntley for an opinion!

[Dmitry Lezhnev @dimalezhnev]

Automated AI can create or copy any software, or SaaS product in hours, days. Completely automated, from idea discovery to publishing, marketing, accepting payments and customers support. That is not "if", that is "when".

[Dmitry Lezhnev @dimalezhnev]

@mattpocockuk cool, that is well formulated, thank you! I'll meditate on that

[Dmitry Lezhnev @dimalezhnev]

@mattpocockuk I was thinking about that, even with all context engineering that I know, in my codebases I can't make it less than 40%, it accumulates: from discovery, to implementation, to verification (a couple of iterations on fixing the code, etc.).

[Dmitry Lezhnev @dimalezhnev]

Founders must fight existing patterns to build a better world.
That takes time.
To be continued.[6/6]

[Dmitry Lezhnev @dimalezhnev]

The initial gap.
A SaaS dev builds a SaaS (part 6) 🧵👇

In media we only see SaaSes that succeeded.
It creates the illusion that founders built something… and users appeared. [1/6]

[Dmitry Lezhnev @dimalezhnev]

This phase is exhausting.
Slow momentum.
Manual work.
Little visible progress.

I think this resistance is why products like this rarely exist.
Habits. Stereotypes. Market inertia.[5/6]

[Dmitry Lezhnev @dimalezhnev]

Two big insights so far:

The platform’s value = rich features + strong builder variety

Cold ads don’t work at this stage. Growth is manual, 1:1, signing people up on their phones as we talk.

More soon.

[Dmitry Lezhnev @dimalezhnev]

# First 50 pieces of content
(A mini-series where a SaaS developer builds a SaaS / 5)
We shipped our MVP, pushed a few ads, and builders immediately started signing up and uploading videos.
[1/4]

[Dmitry Lezhnev @dimalezhnev]

We still don’t know the final shape of the product. We call it a “video portfolio” for now, but that’s not quite right. [3/4]