Cory House

10 lessons I've learned about handling React state over the last 7 years... (thread) #react #reactjs

@samuellawrentz Yes

@housecor But do you read code?

I thought AI would lead to worse code. I was wrong. My code is better than ever because I'm more willing to refactor when AI does it for me. I'm more likely to improve my code when I don't have to worry about how much work it requires. I just sick the clanker on it.

@unfollowedlogic Actually worth a shot

@housecor ask Claude to uncorrupt the DB

My personal site has been on Wordpress for 14 years. Just found out the DB is corrupted and I have no recent backup. Solution: I told Claude "Use data from the Internet Wayback Machine to download every page, and store it as HTML." Now I have static HTML I can rehost!

@glenn_faison Even different stacks can benefit from colocation because I can do an atomic commit for front and backend

@housecor Different stack? I'd rather separate repos. Same stack? I like one repo. I like to colocate everything until it would benefit from being separated

Scenario: Your team supports a web UI and a dedicated backend API. Each uses a different tech stack. Do you put the UI and API in separate repos, or one repo? How do you decide?

@schneidenbach Generally agree but since max context windows recently quadrupled on Opus I’m reevaluating the effects @scottsauber has been experimenting with rarely clearing

@housecor A good rule of thumb is that about 1/8 through the max context window, you should start to think about context management. At a quarter you should definitely clear context

Problem: I'm not sure if I should clear context before my next prompt. Solution: I ask Claude if I should've cleared context. This improves my intuition over time. Example: Prompt 1: "Recover my corrupted blog by pulling files from the Internet Archive." Prompt 2: "Extract each post's content into markdown." Should I have cleared context before step 2? I wasn't sure, so I just asked Claude (I used /btw so it could keep working while answering this side question):

@schneidenbach I agree on point 1, and kept context. And this is the first time I've asked such a thing, but I found the detailed explanation helpful.

@housecor I don’t think keeping context is harmful here. Also, I don’t think an LLM would do a good job of knowing when to clear context tbh. I would not ask Claude/Codex that almost ever lol

@mvidia84853 Yes, I chose to keep the context. But as you can see in the response, it was a gray line whether that was the right call in this case. Tradeoffs to consider.

@housecor Nah , keep the context for step 2. The AI already pulled the blog files in step 1, so it knows what "each post" means. Clearing would make it forget everything and you'd have to repaste it all. Smart to ask tho . builds good habits.

@aminnnn_09 Only around 45 minutes since it had to throttle to avoid getting rate limited

@housecor tbh, that's a genius recovery hack. wayback machine as a backup db is next-level. how long did it take to scrape and convert everything?

@mcbeev @caiobchi Yep top of my list

@housecor @caiobchi Check out Astro.

@avindrafernando Bluehost

@housecor Who were you hosting it through?

@caiobchi Yep that’s the next step

@housecor You could've asked Claude to migrate it to a static site generator like Jekyll or Hugo. That way you could continue to write articles without boilerplate html and in markdown. :)

@ricaldonis @MrPeterLMorris Coloration doesn’t require synced release

Answering to your original question I’d say that it depends. Now, regarding the statement backend changes = UI changes, that’s not necessarily true. There are situations where you definitely want to have those two decoupled from a release perspective. I can provide a few examples if that helps

@JMSchattke Why?

@housecor Every tech stack and compile belongs in its own repository. VMs make hosting multiple repositories an admin, not a purchase, decision.

@walkingriver Yep, and it noticed right away!

@housecor That’s a quick way to discover that the README is broken. I love it.

Last few months I've joined a number of teams. I tell Claude to read the README.md and set up my machine. It does the setup for me, lists issues in the documented setup process, and opens a PR with suggested onboarding doc enhancements. So much easier than by hand.

@avindrafernando Great point!

@housecor Are you assuming the README.md is up to date? I usually ask the agent to do a quick architectural review against what’s in the read me. It’s more work up front but saves a ton of time in case the read me was stale.

@nefthy I struggle to think of a good reason not to sync them when the backend is dedicated to the frontend, and they're in the same repo, and supported by the same team

@housecor Depends on whether deployment is in sync or not.

@MrPeterLMorris Agreed, no guarantee, but easier to assure via automated checks before merge

@housecor Having a single repo wontb guarantee they are in sync. I'd generally keep BeFFe together though.