manton
manton

Not gonna lie, I’m close to dropping $200 to try OpenAI’s Codex. But I don’t think most of my code is well suited to it. Not enough automated tests! It figures that would come back to bite me.

I mostly use AI as a machine that can generate unlimited example code. I learn best from editing examples.

|
Embed
Progress spinner
brandonscript@appdot.net
brandonscript@appdot.net

@manton 100% and even when it gets it wrong, it's still a great rubber duck.

|
Embed
Progress spinner
kylehugh.es
kylehugh.es

@manton It’s very cool, and my bet is that it’s the future of knowledge work (and not for us), but I don’t think it is worth spending that money just to get access to it yet. I have access to it. If you want to be most productive this month, I’d recommend a Claude MAX plan and using Claude Code.

|
Embed
Progress spinner
manton
manton

@brandonscript Exactly.

|
Embed
Progress spinner
manton
manton

@kylehugh.es Thanks, good to know! I should use Claude more.

|
Embed
Progress spinner
danmurrelljr@mastodon.social
danmurrelljr@mastodon.social

@manton If you only have Plus like me, you can try it out with the CLI version. It's pretty interesting as well (you do have to give it an API key though, so will be charged, but it's o4-mini rates)

|
Embed
Progress spinner
jsonbecker
jsonbecker

@manton Claude is pretty good at writing tests…

I am a firm believer that tests tell you what the code should do, and the code itself and docs tells you how it’s accomplishing that goal. When you don’t have tests, you have no idea what the behavior should be. If you don’t have tests, you should be adding them as you work on parts of the code to enshrine your understanding of what you should be accomplishing. Every bug is ultimately undocumented behavior where your users tell you that you were either wrong about what should happen or what should happen was undefined. The bug report itself tells you there’s a part of the should that is in that state, and tests are the best way to document the decision on what should happen and ensure that this behavior is maintained until that decision is changed.

Something I wrote the other day. I will, one day, convince indie developers who spent a lot of time in Objective C to write tests.

|
Embed
Progress spinner
wilgieseler.com
wilgieseler.com

@manton it’s crazy how once you think “i’ll try chatgpt pro for a month” you start relying on it

|
Embed
Progress spinner
In reply to
manton
manton

@jsonbecker It’s a good argument. I’ve known for a while that certain parts of Micro.blog (like feed reading) would be a great fit for tests.

|
Embed
Progress spinner
holgerfrohloff
holgerfrohloff

@manton It’s a regular practice for me now to instruct Cursor to point out which tests I’ve missed and add those. Using these tools you could spend a week and bring your test coverage way up.

|
Embed
Progress spinner
manton
manton

@holgerfrohloff That’s cool. I’m going to be experimenting with this soon.

|
Embed
Progress spinner
holgerfrohloff
holgerfrohloff

@manton probably not going to happen but if at some point you decide to pick up streaming and share those efforts I’d love to see it

|
Embed
Progress spinner