I’ve spent 15 years building software systems. Now I’m building the system that builds systems as the CTO @ Atherio
Joined November 2009
- Tweets 9,308
- Following 268
- Followers 4,386
- Likes 1,611
803 Photos and videos
Dan Patrascu retweeted
In this article, I explore how Cycle-Driven Engineering redistributes many of the responsibilities traditionally carried by code reviews and introduces a framework for deciding which reviews still justify human attention and which not.
danpatrascu.com/blog/why-mos…
1
1
131
Jun 13
One of the most useless takes around newly released coding agents or LLMs are the "one shot" posts.
1
164
Jun 11
I ran 7 different types of workflows based on Fable 5.
Different level of complexity, from quite simple to the most complex of my codebase.
A strange coincidence is that the workflows executed in 12-14 minutes every single time.
Anyone else having similar experiences?
1
170
Jun 11
In my experience, lack of recognition is the fastest path to disengagement at work. Often faster than pay.
Recognition isn't what happens in a private 1:1. It's what others can see.
A short message. A public thank you.
Small intervention. Surprisingly large impact.
2
139
Jun 10
Opus 4.8 every single day.
you can only keep ONE model subscription right now 😭
1) Fable 5
2) Composer 2.5
3) Opus 4.8
4) GPT-5.5
5) Sonnet 4.6
which one are you choosing?
1
203
Dan Patrascu retweeted
Jun 10
I put Fable 5 and Opus 4.8 through the same real-world engineering investigation. The results were not what the hype led me to expect. 👇👇
danpatrascu.com/blog/fable-5…
1
2
3
367
Jun 10
UPDATE: I had some pretty productive and nice results on some more complex parts of Atherio without breaking the bank.
1. main agent -> Fable 5
2. Pinned implementer agents to Opus
3. Pinned reviewers to Fable 5 as well
Basically Fable 5 for "thinking" and Opus for execution
2
146
Jun 9
So it's 2x more expensive than Opus 4.8, but it burns through tokens at around 5x faster. The real cost of Fable 5 is way more than 2x Opus 4.8.
1
1
217
Jun 9
This might be valuable read for anyone looking to get predictable output from coding agents in general and Claude Code in particular danpatrascu.com/blog/my-codi…
5
156
Dan Patrascu retweeted
Jun 9
As a CTO, I like making decisions based on data, not hype. Revisited the AI coding ecosystem after a year and built a synthetic score from measurable signals. Here's the result.
1
1
2
261
Dan Patrascu retweeted
Jun 8
How I Know My System Is Healthy, And What It Costs To Keep It That Way danpatrascu.com/blog/how-i-k…
1
162
Dan Patrascu retweeted
Jun 7
Why AI-Assisted Engineering Fails: The Vibe-Coder's Mindset danpatrascu.com/blog/ai-assi…
2
5
245