Adam @solodevbuilder

🧙🏻‍♀️ AI & Creative-tech leader by day | builder by night 20+ yrs exp leading multi-disciplinary teams and orgs github.com/adam-versed UK Joined August 2023

Tweets

407
Followers

88
Following

335
Likes

153

James Zou @james_y_zou

2 weeks ago

Excited to introduce #Paper2Agent: automatically transform passive paper📜 into active AI agent🤖! Research paper ➡️ paper MCP ➡️ chat agent applying tools/data in the paper to new projects. Key is to use multiple sub-agents to build robust MCP from paper/code. Examples in🧵

15 71 348 33K 341

Download Video

Adam @solodevbuilder

2 weeks ago

im getting neovim pilled - now looking at ultra wide monitors, yes, i am a cliche.

1 0 1 27 0

Adam @solodevbuilder

2 weeks ago

adding a security ci command to the assisted workflows - will analyse your repo for chosen package manager, create a git action that will: - run audits on commits (probably going to change this to have some kind of awareness of where packages have changed), you can also set…

0 0 0 20 0

Adam @solodevbuilder

3 weeks ago

This video on gpt5 gives some good examples comparing the verbosity and reasoning across its model variations youtu.be/MfdgChmZzhg?si…

0 0 0 52 0

Adam @solodevbuilder

3 weeks ago

AI Enablement roles are going to be all over the place over the next 12 months with decent salaries. This is business realising the true value in ai is in augmenting ways of working, more so than using it to develop products. Key word there - augment - not replace

0 0 0 23 0

Adam @solodevbuilder

3 weeks ago

Watching a video supposedly talking about new MCP UI focused tool and one of the opening lines was "LLMS, AI, enabling automatic coding of javascript, iframes.." I properly LOLd.

0 0 0 29 0

Adam @solodevbuilder

3 weeks ago

Codex is a funny one, I'm finding its planning good and i like the way it follows instructions - but i hate the way your are completely abstracted away from the implementation by not being able to access the reasoning and being totally reliant on its summary at the end of tasks…

2 0 0 105 0

Adam @solodevbuilder

4 weeks ago

Giving codex a whirl - claude is extremely buggy and hardwork to use at the moment: - Plan mode generally still guides it well, but theres constant stoppages and hangs that require exiting/resume - It seems to be creating more fall back mechanisms then previously, despite rules…

0 0 0 76 0

Adam @solodevbuilder

4 weeks ago

I've seen different AI enablement roles within industry and one job advert in particular had the following criteria that really jumped out at me: Delivered production agentic AI systems with ≥ 90 % accuracy & < 5 s latency There is an impressive lack of knowledge and…

0 0 0 36 0

Adam @solodevbuilder

4 weeks ago

I'm finding claude code models very finicky of late - if I stick with opus 4.1 (planning mode) and then sonnet 4 for execution for every single task i execute - this actually works well. But if you deviate from that, say continue with a adjacent task with 4.1 enabled - man does…

0 0 0 48 0

Maxime Rivest 🧙‍♂️🦙🐧 @MaximeRivest

a month ago

I just released a 100-minute video on YouTube, divided into three chapters. I’m convinced this video will be valuable to almost anyone coding AI in Python. For beginners in Python and AI, it’s an efficient way to quickly get up to speed. For intermediate and advanced learners,…

6 8 117 7K 124

Adam @solodevbuilder

a month ago

Claude error rate with tool calls seems way higher than normal this last week, there must be some limiting going on for usage because im definitely noticing more failures and dumb actions. Need to get this eval system up and running so i can stop guessing!