Adam Fletcher

I love that all the coding LLMs write a ton of bash, and I also love that they fuck up the quoting 30% of the time. LLMs - they're just like humans.

It's wild how much attribution is held together by utm parameters, and how many tools are awful at including them when needed.

I have said fuck it and given claude access to everything about my business - read only to postgres, email, GA4, posthog, source, logrocket, etc. It's awesome; it generates insights and reports better than I could. Highly recommend.

A great skill in life is knowing how to set the authuser query parameter in google URLs.

Hey @mipsytipsy I love that the Honeycomb install instructions default to using AI and using claude code/cursor with a skill. So cool.

Today: @GeminiApp 503s, @github failing all over the place.

Reboot to patch all the 0-days or keep my claude terminal sessions? The eternal conundrum.

@aye_aye_kaplan @cursor_ai 100% possible that happened - it's a bit hard to know if cursor bugbot autofix is doing anything; I can't find any status anywhere of progress/etc. It'd be neat to see something in app or in-dashboard showing autofix is going (and on what code).

@WriteAheadLog @cursor_ai It looks like in this run, the comments failed to post because the lines couldn't be found. Is it possible you updated the PR like right before Bugbot finished, so we didn't catch it in time to kill the run? Seems like a rare race condition.

@cursor_ai bugbot takes 12 minutes and still requires 3 runs (local cursor deep review doesn't see the same issues), so my pre-merge prompt: "check PR feedback and fix any issues. then, sleep 12 minutes, and check for feedback again and fix, then repeat until no more feedback."

@aye_aye_kaplan @cursor_ai Request ID: serverGenReqId_21dd91bd-4f7c-4a73-b8d3-08108f8768a8 I think - there's a couple of runs in there. Feel free to look at the history in this account if it helps.

@WriteAheadLog @cursor_ai I haven’t ever seen that happen! Can you share a request ID? Should be visible if you click into Cursor Bugbot in the Checks section

Is there a polymarket on which will work first: Apple Intelligence or gmail search?

@aye_aye_kaplan @cursor_ai I turned on Autofix but sometimes it doesn't run and doesn't count the issues correctly; this run actually found 2 issues and the autofix never ran: Cursor Bugbot ... found 3 potential issues. Bugbot Autofix is ON. A Cloud Agent has been kicked off to fix the reported issues.

@WriteAheadLog @cursor_ai We have a feature called Autofix that you can just turn on in your Bugbot settings and it'll do this for you :)

Hi @GMJuditPolgar @Kasparov63 - any chance we can get a much longer Queen of Chess that's edited with a chess player in mind? More game details! More analysis! Also I love you both :)

@aye_aye_kaplan @cursor_ai Yep.

@WriteAheadLog @cursor_ai Are you putting it on Deep mode?

@aye_aye_kaplan @cursor_ai Does it use the same model as I use? Sometimes local deep review will return in 1 min and find nothing, and then via github actions it's 12 mins and 3 bugs, but not all the bugs. I can't _ever_ get deep review to find the same stuff as the action. Nor can anyone on my team.

FWIW local Cursor Deep Review is the *exact same* as Bugbot, there is no difference. The reason you get different results per run is fundamental to LLMs, they are probabilistic and can never guarantee the same results for each run. That being said, this is our #1 priority right now. We are working on getting what we call "consistency" as high as possible to prevent this from happening. We've already mitigated it a lot, but as you can see, it's a hard problem to fully solve :)

@aye_aye_kaplan @cursor_ai I need to set it up, get it to setup my env, etc, etc. But i will!

I don't know what SWR is and I don't know why claude has to keep fixing it in my app, but good for claude.

You're right – this actually *isn't* a place of honor. * I thought esteemed deeds were commemorated here, but that was wrong. * The message wasn't about treasure, it was about danger. That's on me. * I won't sugarcoat it – the danger is still present, and now it's in your body.

@NadimHossain i have been on both sides of this :)

@WriteAheadLog Many such cases (of people who know their shit vs not arguing about something). You could call it Fletcher’s Law

I love the HN threads on datacenters in space. There's always a few actual space engineers who are like "this idea sucks due to space" and then many others who haven't done space arguing that no, MW of heat in space is not a problem and mass isn't real.