Andrew Lisowski 💻
@hipstersmoothie.com
Senior Software Engineeer snowflake.com
Co-Founder @graze.social
Co-Host of @devtools.fm
hipstersmoothie.com
Co-Founder @graze.social
Co-Host of @devtools.fm
hipstersmoothie.com
Reminds me of this
crates.io: Rust Package Registry
crates.io
January 13, 2026 at 3:06 AM
Reminds me of this
Currently name/description/ingredients/tags
January 12, 2026 at 9:20 PM
Currently name/description/ingredients/tags
I would definitely use this. Been thinking out it for sure
January 11, 2026 at 10:25 PM
I would definitely use this. Been thinking out it for sure
Ironically an easing tide would lower all boats Lowering
January 10, 2026 at 6:58 PM
Ironically an easing tide would lower all boats Lowering
A easing tide lifts all boats
Lowering the barrier to entry is a great thing
Lowering the barrier to entry is a great thing
January 10, 2026 at 6:49 PM
A easing tide lifts all boats
Lowering the barrier to entry is a great thing
Lowering the barrier to entry is a great thing
Dan has been saying the same essentially
the hack to being productive with claude code is to start the project with great e2e tests that are declarative enough that they let you precisely specify behavior without saying too much
it's really magic
it's really magic
January 10, 2026 at 6:12 PM
Dan has been saying the same essentially
Every bug fixed without a validation makes your maintenance burden higher for sure
So just make some tests
So just make some tests
January 10, 2026 at 6:10 PM
Every bug fixed without a validation makes your maintenance burden higher for sure
So just make some tests
So just make some tests
You should read the series. It covers all that 😄
January 10, 2026 at 6:07 PM
You should read the series. It covers all that 😄
That’s a good one
January 10, 2026 at 5:13 PM
That’s a good one
I mean being able to run a set of I commands that tells me if my code works as expected
- tests
- lints
- type checks
Like we need some determinism to be confident the thing we’re building still works
- tests
- lints
- type checks
Like we need some determinism to be confident the thing we’re building still works
January 10, 2026 at 5:09 PM
I mean being able to run a set of I commands that tells me if my code works as expected
- tests
- lints
- type checks
Like we need some determinism to be confident the thing we’re building still works
- tests
- lints
- type checks
Like we need some determinism to be confident the thing we’re building still works
Ah it’s an editor kiro.dev
Kiro: Agentic AI development from prototype to production
Kiro helps you do your best work by bringing structure to AI coding with spec-driven development.
kiro.dev
January 10, 2026 at 5:05 PM
Ah it’s an editor kiro.dev
Without unknowingly breaking other things?
January 10, 2026 at 5:04 PM
Without unknowingly breaking other things?
What env doesn’t rely on static analysis to help a person/computer maintain something?
January 10, 2026 at 5:04 PM
What env doesn’t rely on static analysis to help a person/computer maintain something?
That’s not to say the ai might not include that automatically. I could imagine Claude’s system prompts getting tweaked to include stuff like that
January 10, 2026 at 5:02 PM
That’s not to say the ai might not include that automatically. I could imagine Claude’s system prompts getting tweaked to include stuff like that
Would have been such a bigger task to learn both asts and translate it
After that I was a little more of a convert. Before that I hadn’t been able to get anything but tab autocomplete to work well for me
Great tests and solid static analysis seems like a big must
After that I was a little more of a convert. Before that I hadn’t been able to get anything but tab autocomplete to work well for me
Great tests and solid static analysis seems like a big must
January 10, 2026 at 5:00 PM
Would have been such a bigger task to learn both asts and translate it
After that I was a little more of a convert. Before that I hadn’t been able to get anything but tab autocomplete to work well for me
Great tests and solid static analysis seems like a big must
After that I was a little more of a convert. Before that I hadn’t been able to get anything but tab autocomplete to work well for me
Great tests and solid static analysis seems like a big must
I recently wrote an jscodeshift with ai just by defining (+ promoting) tests. Took me about 2 hours to get something pretty full featured
realized I wanted it as an eslint plugin instead and just let the ai port the test suite and fix things. Left it for maybe 30-60 minutes and it just worked
realized I wanted it as an eslint plugin instead and just let the ai port the test suite and fix things. Left it for maybe 30-60 minutes and it just worked
January 10, 2026 at 5:00 PM
I recently wrote an jscodeshift with ai just by defining (+ promoting) tests. Took me about 2 hours to get something pretty full featured
realized I wanted it as an eslint plugin instead and just let the ai port the test suite and fix things. Left it for maybe 30-60 minutes and it just worked
realized I wanted it as an eslint plugin instead and just let the ai port the test suite and fix things. Left it for maybe 30-60 minutes and it just worked
I do think the “pace layers” concept form the series has some merit
Some code can be slower, some faster. Optimizing both for easy regen seems wise. Most of the ways you do that are just by having good maintainable code anyway 🙃 (tests, small files, well defined boundaries, contracts)
Some code can be slower, some faster. Optimizing both for easy regen seems wise. Most of the ways you do that are just by having good maintainable code anyway 🙃 (tests, small files, well defined boundaries, contracts)
January 10, 2026 at 4:53 PM
I do think the “pace layers” concept form the series has some merit
Some code can be slower, some faster. Optimizing both for easy regen seems wise. Most of the ways you do that are just by having good maintainable code anyway 🙃 (tests, small files, well defined boundaries, contracts)
Some code can be slower, some faster. Optimizing both for easy regen seems wise. Most of the ways you do that are just by having good maintainable code anyway 🙃 (tests, small files, well defined boundaries, contracts)
Maintenance is only possible if the have lots of static analysis. I’m not sure the rooses of the world are generating that type of stuff
January 10, 2026 at 4:39 PM
Maintenance is only possible if the have lots of static analysis. I’m not sure the rooses of the world are generating that type of stuff
It’s also a chicken and and egg problem
An artist can look at ai art and see where it’s lacking, a normie can’t do much
Is the same true for software?
An artist can look at ai art and see where it’s lacking, a normie can’t do much
Is the same true for software?
January 10, 2026 at 4:37 PM
It’s also a chicken and and egg problem
An artist can look at ai art and see where it’s lacking, a normie can’t do much
Is the same true for software?
An artist can look at ai art and see where it’s lacking, a normie can’t do much
Is the same true for software?