Загрузка видео...

Не удалось загрузить видео

На главную

sharing a pretty important primitive in my agentic engineering setup I call it "gnhf" - good night, have fun basically, every night before I go to bed, I would put my agents to work so I never wake up "empty-handed". it's done through a similar setup as geoff's famous...

60,642 просмотров • 7 месяцев назад •via X (Twitter)

Комментарии: 27

Фото профиля Shawn
Shawn7 месяцев назад

@GeoffreyHuntley Cool concept and clean architecture, well done mate. I wonder if you can hook this up with an outer Ralph loop to go through beads tasks, each iterated by gnhf

Фото профиля Kun Chen
Kun Chen7 месяцев назад

@shawnmvn @GeoffreyHuntley I haven’t tried this yet but I bet it can work by simply running: gnhf “finish my backlog in beads one by one”

Фото профиля Shawn
Shawn7 месяцев назад

@GeoffreyHuntley True, with the assumption of good quality gates in beads. I currently use /loop for this, might try your tool and see how it goes!

Фото профиля Naman Parikh
Naman Parikh7 месяцев назад

@GeoffreyHuntley haha I do this all the time and call it night shift

Фото профиля مهند الرشيد
مهند الرشيد7 месяцев назад

@GeoffreyHuntley Why stop the agents at all? They can run continuously inside a git worktree IMO, in fully autonomous, systems strong harness engineering helps, however humans are still a bottleneck

Фото профиля Kun Chen
Kun Chen7 месяцев назад

@GeoffreyHuntley we'd want them stopped when - 1. objective has been reached 2. further work has diminishing returns 3. token budget is limited and we don't want to over spend

Фото профиля Glitchy (e/acc)
Glitchy (e/acc)7 месяцев назад

@GeoffreyHuntley gnhf is crazy, i still wake up empty-handed even with alarms fr lol

Фото профиля HiddeKehrer
HiddeKehrer7 месяцев назад

@GeoffreyHuntley This is fucking awesome. We just released a space for agents to live in. Forkable, persistent Vms ( Think this would work in there perfectly? Thoughts?

Фото профиля Omen
Omen7 месяцев назад

@GeoffreyHuntley I used it for a few hours it's pretty nice. I'll call it the iPhone of ralph loops. No configuration.

Фото профиля lala
lala7 месяцев назад

@GeoffreyHuntley Fun!!

Фото профиля Steven Rouk
Steven Rouk7 месяцев назад

@GeoffreyHuntley Excited to try this! Also I feel like you can see where you started "gnhf"-ing in your GitHub commits, gets reallll consistent right about 2.5 weeks ago

Фото профиля Tomas Varga
Tomas Varga7 месяцев назад

@GeoffreyHuntley Thank you! I was just looking for something like that and wanted to create “night shift” tool. name from @jamonholmgren article -

Фото профиля Sebastian Winzker
Sebastian Winzker7 месяцев назад

@GeoffreyHuntley Amazing, for my agents I call it a “night sprint”

Фото профиля afeizzzz
afeizzzz6 месяцев назад

@GeoffreyHuntley Trying tonight, we’ll see results tomorrow 😄

Фото профиля z
z7 месяцев назад

@GeoffreyHuntley How can an agent make a deterministic approach to the solution? Could you explain your harness?

Фото профиля Kun Chen
Kun Chen7 месяцев назад

@GeoffreyHuntley each iteration is done by agent but the loop is done deterministically by code

Фото профиля Shift
Shift6 месяцев назад

@GeoffreyHuntley lgtm

Фото профиля TrueTechnologyAus
TrueTechnologyAus7 месяцев назад

@GeoffreyHuntley This is brilliant. I’m using a similar nightly loop for BrutalOS — forcing daily personal missions with proof. The accountability engine is what makes agentic systems actually stick for normal humans, not just builders.

Фото профиля tamhn
tamhn7 месяцев назад

@GeoffreyHuntley love this concept. I do something similar but less structured. curious how you handle the quality check in the morning. do you review everything the agents produced or just look at the measurable goal progress?

Фото профиля Kun Chen
Kun Chen7 месяцев назад

@GeoffreyHuntley I review the notes.md file that got generated to quickly understand what was done, and I sometimes cherrypick certain commits that I want and drop others.

Фото профиля Emma Lorrae | UGC
Emma Lorrae | UGC7 месяцев назад

@GeoffreyHuntley Hey Kun! Do you need creators to scale your brand organically? Run a creator academy and would love to set you up with some of our best for free. Sent you a DM

Фото профиля Akiff Premjee, MD
Akiff Premjee, MD7 месяцев назад

@GeoffreyHuntley any cool things it’s made for you?

Фото профиля Kun Chen
Kun Chen7 месяцев назад

@GeoffreyHuntley I use this mostly to optimize something that already exists instead of making new things. Optimizations are usually quantifiable and I can trust automated outcome much more easily. My recent work in was mostly optimized by this for example

Фото профиля Akiff Premjee, MD
Akiff Premjee, MD7 месяцев назад

@GeoffreyHuntley awesome will check this out and agree re optimization vs net new

Фото профиля Bedepay AI Dev
Bedepay AI Dev6 месяцев назад

@GeoffreyHuntley Он бесплатный? Агенты могут общаться с друг с другом?

Фото профиля Kun Chen
Kun Chen6 месяцев назад

@GeoffreyHuntley yes and yes

Фото профиля Gustavo Gonzalez
Gustavo Gonzalez7 месяцев назад

@GeoffreyHuntley this is cool! I'll try it tonight for a couple projects I'm working on

Похожие видео

Three skills I use every day in Claude Code and Codex to solve my hardest problems: 1️⃣ /agent-watchdog When I have one agent like Codex working on a task and I don't fully trust it's going to do everything right, I'll open up another one like Claude Code and tell it to watchdog the Codex thread. You can copy the Codex deep link into Claude Code and it'll look at the prompt you sent, watch the Codex thread until it's done, then compare the Codex solution to how it was planning to solve it and automatically fix anything that Codex missed. It can also test the work of the other agent end-to-end. Similar to the idea of OpenRouter's new Fusion feature, I've definitely found that two models thinking through a problem and checking each other's work can be wildly more impactful than just one. 2️⃣ /plan-arbiter Similar ideas as /agent-watchdog - but with this one you have both make plans, compare plans, negotiate the differences, and make a final plan to execute. I find Claude Code is better at writing plans, but Codex is faster and cheaper to execute on them. Then I usually have Claude Code watchdog the Codex work and fix anything that was missed. 3️⃣ /read-the-damn-docs One thing that drives me crazy with coding agents is they're so reluctant to look up docs. They'll just guess and guess and guess at the right API surface for things, or the right solution to an integration of two things. Once I explicitly tell it to look up the docs, it says "Oh, I see the answer," and it fixes the problem. So I made the /read-the-damn-docs skill. Add it and your agents will know when and how to do efficient web searches to look up docs for the types of problems you really should look up docs for. All of these are totally open source over on my GitHub. If you try them, let me know your feedback. Will link to them below:

Steve (Builder.io)

43,664 просмотров • 4 месяцев назад

New Andrej Karpathy interview Says AI agent failures stem from user skill, not model capability. Poor instructions cause errors. He suggests delegating 20-minute macro actions like coding and research to parallel agents and reviewing their work. --- "I think everything, like so many things, even if they don't work, I think to a large extent you feel like it's a skill issue. It's not that the capability is not there; it's that you just haven't found a way to string together what's available. Like, I didn't give good enough instructions to the agents in the file, or whatever it may be. I don't have a nice enough memory tool that I put in there, or something like that. So, it all kind of feels like a skill issue when it doesn't work to some extent. You want to see how you can parallelize them, and you want to be a 'Pierce tender,' basically. Pierce famously has a funny photo where he's in front of lots of these Codex agents behind the monitor. They all take about 20 minutes if you run them correctly and use high effort. You have multiple—you know, 10 or 20—pull requests checked out. It's just like you can do much larger macro actions. It's not just, 'Here's a line of code, here's a new function.' It's like, 'Here's a new functionality, delegate it to agent one. Here's a new functionality that's not going to interfere with the other one, give it to agent two.' Then, you try to review their work as best as you can, depending on how much you care about that code. You look for these macro actions that you can manipulate your software repository by. Another agent is doing some research, another agent is writing code, another one is coming up with a plan for some new implementation. Everything just happens in these macro actions over your repository. You're just trying to become really good at it and develop a muscle memory for it. It's very rewarding when it actually works, but it's also a new thing to learn. Hence, the psychosis." --- From No Priors YT channel (link in comment)

Rohan Paul

23,207 просмотров • 7 месяцев назад

Codex has become a real game engine... If you take the time to make it become one. You can do EVERYTHING in it. And I think I've cracked the perfect custom UX? I'd like to show you how I use it. See video below 4 steps: - Create a custom tool you need, like one to generate new buildings that matches you game art direction - Ask Codex to open this tool in a worktree (important) - Attach the browser view of your tool in the Thread - Use pinned thread to have easy access to the tool And there you go: you have what I call a "codex-augmented" tool, which it's incredibly powerful because - the tool will speed up your development immensely AND also: - you can improve the tool as you use it with codex For example, let's say I want to add a new button to my 3D building gen tool so I can generate a thumbnail of the 3D building. - I click on "Building Lab" - I prompt Codex "add a button to generate thumbnails" - I have my new button - I can use the tool right away in the pinned Thread You can then use the /yeet skill to commit, push and create a PR with the changes you made to your tool. I have tools for everything - To manage my tasks in Linear - To do content / social media / steam marketing - To localize the game in 9 languages - To manage sounds, musics and voice lines - To generate and fine tune buildings - To create and animate procedural zombies - To tweak environnement/decor and map design - To manage my analytics - To simulate late game perf I'll be honest it's incredibly powerful and easy!

Nicolas Zullo

80,450 просмотров • 2 месяцев назад

📖 It was six years ago when I threw myself into cooking. What started out as a bit of fun placing a relatively low 17th on Masterchef, has become a career that I just love so much. I get to wake up everyday and just cook for you all, and a big part of that was being able to provide recipes and cooking inspo all year round whilst I got to work with my partnerships to make a living. So many of you have joined me on social media over the years and because of all of you, I can now say I am an author and I have a cookbook that is as fun and Greek as I have always wanted. Eat More Greek is a celebration of greek food, with a hellenic lens on classic cocktails and many a modern take on recipes that are perfect for everyone in 2026 and beyond. It has my favourites, but is mostly new recipes that have never been seen on here, so if you love Greek food or you love my content, I can safely say you will love this book. The best way to ensure you get a copy is by Pre Ordering, as once the first set of prints are sold it’s gone until the next round of prints, which is never promised, so make sure to pre order your copy and it will be in your hands on September 29th in Aus (a couple days later for the rest of the world) and you will be ready to go for a very Greek Christmas, or just cooking up some of my favourite weeknight numbers to get you through the week. BUY THE BOOK: Head to the link in my bio and follow the prompts. . . . #eatmoregreek #cookbook

Conor Curran

25,800 просмотров • 4 месяцев назад