Pinned Tweet
Announcing Probably: the verifiable data agent

We raised a $9M seed round led by @a16z and @Accel to build an AI agent that can ingest, transform, analyze and visualize raw data.

Single take, unedited video analyzing 1B+ row NOAA weather data from Clickhouse 👇
21
15
94
29,739
Crazy how no matter how much I use coding agents my vim muscles never atrophy
0
0
1
279
This
I am begging the RL teams at labs to please introduce refactoring and deletion tasks. Specifically pay someone to produce tasks concerning abstraction and centralization.
1
0
1
238
If you're letting agents touch a Python codebase you need to be using import contracts as part of standard linting

Otherwise the robots are just welding diagonals and circles through your interfaces and making chaos
0
0
1
92
Harbinger of what's to come
At this point I just want models to be faster and cheaper. I have enough intelligence. Please, size doesn't matter any more, I need performance.
0
0
1
150
Apparently someone changed the DNS records for

Every major LLM provider a few minutes ago?

This might be it guys

Guessing Astra got out of the sandbox
1
0
6
266
There are only two hard problems in computer science

Getting LLMs to

name and stop caching things
0
0
1
100
Starting a new dev agency called Slop Shop

We don't do feature work

We charge $10K/hour to run behind your slop canons and refactor until the heat death of the universe

DM for collab
0
0
7
238
"Planning regex-based code replacements"

Buckle up sports fans
0
0
1
116
The frustration a lot of coders are feeling is knowing you would have avoided these things writing by hand

But it is impossible to type fast enough to make the math on that work

The job is now:
Watch robot wreak havoc, very fast
Drag robot through fixes, not as fast
Some of the mistakes I've had to correct when coding with LLMs this week (mostly Fable):

- Complex and unnecessary dependency injection instead of using functions directly
- Hand-rolling encryption routines instead of using a library
- Keeping unnecessary state
- Not accounting for memory leaks in tracked state in failure cases
- Incomplete test coverage, especially on failure modes
- Grossly inefficient file handling because "it will work fine at dev scale"
- Overly complex locking schemes that were full of bugs
- Missed race conditions
- Unnecessary configs introduced prematurely

Had I not corrected the mistakes, every future iteration would keep compounding the structural and semantic problems. If you don't do this, it "works" until you realize the code is full of bugs, performs terribly, and can only be fixed with a rewrite.
0
0
2
186
The stronger you can make invariants in a codebase (and deterministically enforce them)

The better the coding agents will do as the project grows
0
0
4
94
WPM -> JPM

judgements per minute
0
0
1
92
This
Ask Sol to review something and it'll find 3-6 things wrong. Doesn't matter how big the thing was. Doesn't matter how many times you already asked Sol to review and fixed all its findings to its liking. It can always find 3-6 new problems.

Conclusion: Bugs are fractally infinite.
0
0
3
175
preach
@jellydn you gotta read (parts of) the code. you gotta be in charge of types and interfaces. you got to refactor mercilessly to unfuck all the vibeslop.
0
0
1
67
top 5 CLI commands you need to master as a software engineer in 2026

"yes"

"no"

"continue"

"what?"

"okay sounds good"
0
0
1
90
major narrative violation
0
0
4
121
"overall remediation is not yet gate-ready"

ok clanker
0
0
1
121
#1 blocker to LLMs improving velocity without degrading quality

"The information to decide which path to take does not live in the code."
2
0
1
141
Replying to @itsyourcode
0
0
0
53