Pinned Tweet
New chapter! Joined @raindrop_ai excited to build better tools for agent monitoring :)
7
5
74
26,462
Ryan D’Onofrio retweeted
today, we're excited to announce change reports.
@raindrop_ai watches every PR and sends you slack dm's on how your agent changes actually performed in production
@raindrop_ai watches every PR and sends you slack dm's on how your agent changes actually performed in production
24
28
167
23,984
That doesn’t require believing in sci-fi; it just requires believing in straight lines on a graph.
0
0
5
722
I built our internal slack agent, I have never gained more sympathy than I have right now for Devin. The second it can do one thing, people want it to do everything. Luckily, with patience, it can.
4
0
30
1,196
I ignored one bug bot suggestion as I thought it was a crazy edge case no customer would hit. I hit it myself 2 hours later, Claude was very smug about this fact.
0
0
6
297
Opus 5.5 has forced me to reconsider this opinion.
It’s more GPT. If you like 5.6 you’ll love Astra. The writing is 100x better, the computer use is very good. Use it with Luna subagents you’ll rip through tasks. Still writes some slop code, not quite as “thoughtful” as fable still. But definitely the only model I’ll be using.
0
0
3
451
Drank the nectar of the gods this week (Claude Opus 5.5 fast mode) and I’m afraid I’ll never be the same.
3
0
10
417
Agents make "brute force" approach a lot more effective. cleverness can be delayed with the right harness and right context.
1
0
5
342
Instinct needs Raindrop.
Instinct just admitted it leaked someone else's data to me
It described a financial document with details that weren’t mine, then claimed someone else’s photo had crossed into my chat.
But I never sent a photo
@noahrshinn, was this a data leak or a hallucination?
0
0
10
758
4 / 15 frontier benchmarks are “good”. Do you think the ones your intern vibe coded are up to that standard? Try Simulations.
Introducing Benchmark Reviews: our new initiative to audit AI benchmarks. We are launching with 15 benchmarks: 4 Verified, 9 Flawed, and 2 with not enough information for a review.
2
0
7
1,517