Pinned Tweet
In huge markets such as K-12, there's data everywhere. AI agents are simply not enough to prospect and outreach efficiently at scale.
2
0
8
108
yo @thsottiaux codex is egregiously slow. Astra on 1.5x speed is insanely slow. If this is the new normal I am so out on OpenAI models.
1
0
0
33
@thsottiaux dont even want a reset, just need my tasks to get done this century
0
0
0
12
in search of a cracked engineer. Must be in NYC or willing to come build here. Funded stealth startup. Dm me
0
0
2
47
Astra actually really sucks right now what is going on, it's making very elementary searching mistakes that we haven't seen since ~2024
0
0
0
23
Quickly my local mac is becoming a bottleneck for development speed. Worktrees cluttered everywhere, disk space problems, RAM not enough to keep up with the number of agents running.
Has anyone successfully done a Remote Development Environment or do I need to buy a better computer?
Has anyone successfully done a Remote Development Environment or do I need to buy a better computer?
1
0
1
17
I've noticed that Claude (in CC) will delegate tasks to sub-agents and not read the code or take responsibility for what bad decisions were produced. Sub-agents in this way feel worse than having claude directly build within the parent chat.
anyone else notice this
anyone else notice this
2
0
3
174
While i'm complaining:
This subagent model (especially in CC) ends up with features that the parent agent has never even known about.
I'm arguing with a model that has never even seed or heard of the implementation it's defending. It's maddening!
This subagent model (especially in CC) ends up with features that the parent agent has never even known about.
I'm arguing with a model that has never even seed or heard of the implementation it's defending. It's maddening!
0
0
0
14
I know this is old news by this point but my experience on Codex is vastly better to that in Claude Code.
Opus 5 is just unusable and annoying to talk to, and Sol just churns out goals like it's nothing. Incredible software and what a remarkable turnaround.
Opus 5 is just unusable and annoying to talk to, and Sol just churns out goals like it's nothing. Incredible software and what a remarkable turnaround.
0
0
0
50
I have no idea what's going on but it cannot figure out very simple and basic tasks that I have been comfortably asking AI to do for > year.
0
0
1
21
has anyone figured out how to stop refactoring regressions in agents? Potentially skill issue but every time I try and bring functionality from one place to another with an agent, it can never get it right / leaves something big out. Huge observability problem
0
0
0
19
Has anyone figured out how to manage db migrations while agents touch new features? have a jank-ass setup for this currently
1
0
0
12