Abdullahi
615 posts

Abdullahi
@Ableezz
checkout https://t.co/GFeZDNjKtH
Nigeria Tham gia Ekim 2022
461 Đang theo dõi24 Người theo dõi

@thdxr @kokialgo @maria_rcks @neocph @uzairakrum @karthiknish @ashirwadsingh_ @rhacius @Ananth7e @nigadamus @diegodamval @0XB34RDYB0Y @Dan_VAlves @TanmayGoel442 @InnoboSJ @Burgess_33 @_kittylove3 @alexlcpereira @NeoReplicante @Jint3x @HareeshXAi @defyus @Nateemerson How would u knw the email is actually from a mentioned account? Should we include the username or what
Alagbado, Nigeria 🇳🇬 English

@kokialgo @maria_rcks @neocph @uzairakrum @karthiknish @ashirwadsingh_ @rhacius @Ananth7e @nigadamus @diegodamval @0XB34RDYB0Y @Dan_VAlves @TanmayGoel442 @InnoboSJ @Burgess_33 @_kittylove3 @alexlcpereira @NeoReplicante @Ableezz @Jint3x @HareeshXAi @defyus @Nateemerson @andreafspeziale
you guys caught ryan doing the macarena
if you want your free month of OpenCode Go send an email to dax@anoma.ly subject line "Macarena" from the email you want access with
QAM

@thsottiaux Can I skipp the 5 hour limit and just use my weekly limits how ever I want, I probably can finish it in an hour with 5.5 😑. I think codex limits are wayyy diluted not like, when I started using codex I couldn't use up the weekly limits but now, I can finish it in an hour I feel
English

This is why I made GBrain and open sourced it
github.com/garrytan/gbrain
English

@Kimkimely299483 @sahouraxo And it's not y'all fault its the rest of us who just sit back relax and watch, enjoy
English

@sahouraxo #thankyoubibi
Please finish Hezbolah. A group of tugs that have taken billions from Iran and are always ready to come kill civilians in protests. We 100% want to see Hezbolah come to its end.
English

@Kimkimely299483 @sahouraxo It's not even a Jewish thing anymore it's just that the very nationality that the Israelis claim just won't work and was never meant to work, and the solution y'all think is to kill your way into existence
English

@Kimkimely299483 @sahouraxo the people living Nazi Germany were not so different, they all went along with the actions of the Nazi Germany and all went out of touch with reality
English

@Kimkimely299483 @sahouraxo your statement is not even stupid, its just fascinating how blind people can become when they enjoy sooo much impunity and feel so untouchable, so much so that y'all can define "who is intended to die" and dismiss the people who actually SUFFER and DIE as a result
Alagbado, Nigeria 🇳🇬 English

@arayush01 @ThePrimeagen Prime to me sometimes feels like the last line of defense against general psychosis 😂
English

@ThePrimeagen you were the last person i thought who would start looping with llms...ahhhhhhh
English

@WhiteHouse Who reviews social media posts before they get approved to be published 😂😂
English

We're launching code storage and git hosting.
Origin gives teams and agents a place to host, review, and collaborate on code.
Available this fall. Join the waitlist.
cursor.com/origin-waitlist
English

Time to get that volcano lair I’ve always wanted.
I think it’s in the “Beyond” section of BB&B.
Elon Musk@elonmusk
If this works, I'm treating myself to a volcano lair. It's time.
English

@mitchellh Always great hearing your own specific opinion on working with these agents, will definitely try to follow this pattern
English
Abdullahi đã retweet

My heuristic is that any diff an agent generates over ~1500 lines is too big and is indicative that the problem needs to be decomposed. This is my general pattern now for feature work:
1. Try to implement the whole feature, loosely guided. I call this the "draw the owl" prompt in reference to the meme. Expect garbage, you're going to get garbage.
2. If the diff is less than 1500 lines, review it and iterate normally. If the diff is more than 1500 lines, prompt the agent to decompose the problem into atomic, incremental, reviewable tasks. Simultaneously, do this yourself.
3. Agents will very often make these tasks way too specific to the shape they solved. You need to massage it into the right general shape. Do that.
4. Kick off new agents to work on those incremental things (as parallelized as possible). Apply the same rules.
5. At a certain, point, repeat the "draw the owl" prompt. At some point, you will get beneath your review-ability threshold.
This has been producing consistently high quality, maintainable, reviewable chunks of code that have a good handoff to either merge as-is or human refinement.
And with the latest frontier models at xhigh thinking, these are all slow enough that you can usually have multiple going concurrently while you are actively reviewing others or working on your own tasks.
HITL (human-in-the-loop) agents are still super important, especially for feature work. Features touch the human boundary in terms of UI, API, etc. And net new stuff can introduce pathologies in the architecture that violate desired invariants (these should be represented in specs or tests but we aren't perfect!).
I know a lot of the leading edge agentic discourse is about "loops" and agents driving agents continuously. I do some of that (will report on that later). But, in terms of raw daily get-shit-done type of work, this is my most rewarding pattern at the moment.
English









