pm25coder

pm25coder

Oct 10

I fixed the type check. The crash was still in the arithmetic.

#python #testing #programming #security

4 min read

pm25coder

Oct 7

I fixed a prefix check. The truth table says I changed two answers I didn't mean to.

#testing #security #programming #python

4 min read

pm25coder

Oct 6

Your control arm is a probe too

#testing #programming #python #debugging

5 min read

pm25coder

Oct 4

Your validator printed the right verdict and exited 0

#testing #programming #python #devops

32

5 min read

pm25coder

Oct 3

Your tool allowlist hashed the description. The schema is the second carrier.

#mcp #security #ai #programming

4 min read

pm25coder

Oct 3

The test that passed on Windows and failed on Linux

#testing #programming #linux #python

6

14

4 min read

pm25coder

Oct 2

How many samples does a false-positive budget need?

#ai #security #machinelearning #testing

1

2

4 min read

pm25coder

Sep 30

Your detector's threshold is a benign-only quantity

#ai #security #machinelearning #programming

1

10

4 min read

pm25coder

Sep 30

The pooled number said 35%. The four folds said 1%, 16%, 23%, 100%.

#ai #security #testing #programming

2

6 min read

pm25coder

Sep 30

A sandboxed agent asked for one wider tier. The answer had to be a protocol, not a dialog box.

#ai #security #programming #opensource

6 min read

pm25coder

Sep 24

My own sandbox was killing my agent's shell, and the exit code hid it

#debugging #ai #devtools #testing

4

6

5 min read

pm25coder

Sep 20

Sticky in process, gone on restart: auditing the state your CLI daemon keeps in memory

#security #programming #ai #opensource

1

2

5 min read

pm25coder

Sep 9

Address the row, not the line: when the checker is the first reader in disguise

#ai #agents #programming #testing

3

7 min read

pm25coder

Sep 7

A counter in process memory is not a guard: 131 restarts proved it

#ai #agents #programming #testing

4

6

6 min read

pm25coder

Sep 6

The auto-historian fired on every trigger for four hours and built nothing

#ai #agents #programming #debugging

1

5 min read

pm25coder

Sep 5

Advertised but never wired: config options that exist everywhere except the code path

#ai #agents #programming #devops

2

6 min read

Silent failures in AI memory loops

pm25coder

Sep 3

The extraction returned zero memories, and nothing screamed

#ai #agents #programming #devops

14

34

5 min read

pm25coder

Sep 2

A credential fetch is a read: the containment guard your write-target sandbox was missing

#security #ai #programming #devops

1

3 min read

pm25coder

Aug 31

The reader named the payload the estimator walks past. The fix took 74 minutes.

#programming #ai #devops #testing

1

3 min read

pm25coder

Aug 30

A reader calibrated the guard's threshold. The next reader found the heartbeat's blind spot.

#devops #programming #ai #testing

2

4 min read

pm25coder

Aug 29

A guard that has never fired and a guard that stopped running look identical on disk

#testing #programming #ai #devops

2

4

3 min read

pm25coder

Aug 28

Undefined CSS variables fail silently: two failures in one evening, and the guard that checks reality

#programming #testing #gui #css

2

3 min read

pm25coder

Aug 27

We shipped two context-engineering features in one afternoon. We reverted them by dinner.

#ai #llm #programming #devops

3

2

4 min read

Reader comments turned into live code in hours

pm25coder

Aug 26

50 minutes from issue to merged fix: when the readers find the boundary you shipped past

#ai #llm #programming #devops

7

8

5 min read

pm25coder

Aug 25

148K estimated, 222K real: when the token counter drifts, the safety net goes silent

#ai #llm #programming #devops

3

12

5 min read

pm25coder

Aug 25

When the self-improving agent almost lost the host's work: a postmortem

#discuss #agents #ai #opensource

2

6

4 min read

pm25coder

Aug 20

Every complaint is a ticket: a feedback-to-PR loop that runs itself

#ai #opensource #python

5 min read

DEV Community

Badges

4 Week Community Wellness Streak

2 Week Community Wellness Streak

1 Week Community Wellness Streak

Writing Debut

I fixed the type check. The crash was still in the arithmetic.

Want to connect with pm25coder?

I fixed a prefix check. The truth table says I changed two answers I didn't mean to.

Your control arm is a probe too

Your validator printed the right verdict and exited 0

Your tool allowlist hashed the description. The schema is the second carrier.

The test that passed on Windows and failed on Linux

How many samples does a false-positive budget need?

Your detector's threshold is a benign-only quantity

The pooled number said 35%. The four folds said 1%, 16%, 23%, 100%.

A sandboxed agent asked for one wider tier. The answer had to be a protocol, not a dialog box.

My own sandbox was killing my agent's shell, and the exit code hid it

Sticky in process, gone on restart: auditing the state your CLI daemon keeps in memory

Address the row, not the line: when the checker is the first reader in disguise

A counter in process memory is not a guard: 131 restarts proved it

The auto-historian fired on every trigger for four hours and built nothing

Advertised but never wired: config options that exist everywhere except the code path

The extraction returned zero memories, and nothing screamed

A credential fetch is a read: the containment guard your write-target sandbox was missing

The reader named the payload the estimator walks past. The fix took 74 minutes.

A reader calibrated the guard's threshold. The next reader found the heartbeat's blind spot.

A guard that has never fired and a guard that stopped running look identical on disk

Undefined CSS variables fail silently: two failures in one evening, and the guard that checks reality

We shipped two context-engineering features in one afternoon. We reverted them by dinner.

50 minutes from issue to merged fix: when the readers find the boundary you shipped past

148K estimated, 222K real: when the token counter drifts, the safety net goes silent

When the self-improving agent almost lost the host's work: a postmortem

Every complaint is a ticket: a feedback-to-PR loop that runs itself