Andrei Maxim

This account is verified. Learn more

@rockatanescu

Junior applied AI engineer, ex-Ruby developer

Bucharest, Romania andreimaxim.com Joined February 2022

432 Following 763 Followers 5K Posts

Latest posts 5

Andrei Maxim @rockatanescu 3d

I don't understand how these posts get 1.6K likes in November 2026.

This is the Ralph loop workflow circa January 2026 and it only builds a slop cannon.

Trackmind Verified account @0xTrackmind 6d

this harness engineering paper is f*cking insane

it's the exact setup anthropic used to keep a single AI agent coding for hours without forgetting what it was doing.

2 agents, 4 files, one prompt that runs every session, all on one page.

set it up tonight and your agent keeps building feature after feature while you sleep, instead of quitting halfway.

the crazy part is everyone blames the model when their agent gives up after 20 minutes, while the fix is a json file and a text log.

full breakdown in the harness guide below.

View quoted post on X

Andrei Maxim @rockatanescu 4d

Amazon researchers recently published a paper which I think is notable considering that Amazon has put a lot of effort into spec-driven development.

The paper argues that specifying how a model should implement a specific solution is significantly superior to writing prose.

Andrei Maxim @rockatanescu 6d

Looking at these testimonials, I don't see any acceleration beyond what you wouldn't expect to see in a regular software company that's using the currently available models.

It's also interesting that Opus 4.5 was such an inflection point even internally at Anthropic.

elie Verified account @eliebakouch Nov 5

i think it's clear that the role of ai in AI R&D went from autonomous coding to autonomous experimentation hence reducing the time to experiment by a crazy amount. from what i see in the interviews we are not yet at the stage where models are also used to autonomously generate new research ideas

the big issue is that they don't release any metrics that would allow us to quantify this in this report or in any system card..

View quoted post on X

Andrei Maxim @rockatanescu 6d

Lots of things to do and learn.

Cindy Sridharan @copyconstruct 6d

@poteto A mix of the techniques shown here, in short.

This has been my dream for nearly a decade now, and with agents I’m finally able to see this (partially) come true.

But turns out this is what I necessarily require to trust agent generated code. It’s not optional anymore, imo.

View quoted post on X

Andrei Maxim @rockatanescu Nov 5

Great thread.

Jack Ellis Verified account @JackEllis Nov 4

I'm not convinced that senior software engineers are pushing code without reviewing it in some way. So I'm going to tag eleven experienced engineers in this thread, who are building high stakes software, to see what they do. Let's see what we learn.
View quoted post on X