Griff @Griffware_

PHP warns of another unpatched RCE Bernie4Life420, a philosopher:

seems like every iOS app developer should be doing this 1. IDENTIFY THE EMOTIONAL PEAKS AND THE MOST VALUABLE MOMENTS FOR YOUR USER - in my app, @miji_app this is when you finish an awesome practice session 2. ASK FOR A REVIEW RIGHT THEN -- AT THE USER'S EMOTIONAL PEAK

As long as you're deliberate about WHY you're willing to engage in the probabilistic lottery, that's fine. If you're out of your expertise, and you're ok with a likely answer, then cool. But if you want something specific, you must be deliberately specific.

For example, in 2016, the way to be deliberate in coding was to, well, code a function as thoughtfully as you could. Want a good sorting algorithm? Implement the best combination of elements from different algorithms, line by line. Be deliberate. It's still the same today, except you can just give your agent context about what you want. Today, you have to be deliberate about exactly what you want, and the agent will fill in the details. Is quicksort best? Not sure, maybe it's bubble sort, or maybe it's... a combination of elements from different algorithms, tuned to the context we're working with. if you want a specific implementation, ask for it. Otherwise, it'll take a stab, and you iterate based on the results.

The MOST important skill in agentic coding is: BEING DELIBERATE. The less deliberate you are, the more you leave the implementation details up to raw probability. That's fine, if you don't mind. But for scenarios that require taste, you NEED to be thoroughly deliberate.

Finally got approved for the iOS app store! First app published :D I actually ship things!

This is the most important skill in AI and you’re trolling if you don’t do this Building an autonomous YouTube content production pipeline with Codex / Claude Most prompts I write nowadays are basically entire essays. You need to be able to express your vision as clearly and thoroughly as possible. Remember: anything you don’t mention WILL be guessed at. You need to be deliberate with where you want to leverage the probabilistic nature of these tools, and where you want to restrict it to your exact philosophy / reasoning / framework. 3099 words is just one prompt of hundreds I’ve sent for building this tool. That’s what it takes. Don’t outsource your mind, but do leverage the tool to implement YOUR vision. Anything unmentioned is a guess.

what if indie game devs could provide all of their game assets -> apply these same advertising formats? basically re-leveraging all the time and love and care put into your models / art --> doing all the boring marketing stuff

Higgsfield AI 🧩 @higgsfield_ai

3 weeks ago

Introducing Ad Multiplier in Ads Studio. Turn any winning ad into data-backed creative for your brand. > Decodes what drives its performance > Applies those insights to your product > Generates new ads ready to test and scale Live on Higgsfield and in the ChatGPT extension.

94 275 1K 323K 850

Closed source is definitely not dead For example, how are you going to decompile the TikTok Algorithm? How Amazon’s marketplace ranks products? There’s a lot of closed source software that doesn’t (and will not ever) exist on your machine. I’m being pedantic; obviously the point is that for software that actually runs entirely on your hardware, yes, it can be decompiled and yoinked. Surely this will be good for competition and it won’t just result in more draconian laws, right?

Eric S. Raymond @esrtweet

3 weeks ago

This is the doom I predicted a few days ago, coming for Photoshop. A clean-room open-source reimplementation. No prizes for guessing that they decompiled Photoshop to source code, processed that to some kind of non-code specification language, then fed the spec to an LLM with an

635 1K 17K 4.0M 16K

This is the truth in 2026 and beyond Ex-TikTok MLE / Ex-Amazon SWE; I've been coding since 2016. Especially when leveraged via agent swarms, we're now able to build things that were only dreams with AI agents -- because creating certain tools (like Rollback) takes a godly amount of attention and sustained effort AND god tier skills. Now it takes less skill to implement, but the skill shifted towards how to manage the swarms, how to make things not slop, etc. It's shifted towards your vision, WHY you're building, and building the right things. If anything can be built, why are we building? That's the challenge. I've been building github.com/griffinbeels/S… over the past couple of months, with over 2000+ commits. I've spent easily hundreds of hours on this project, but I haven't coded at all; my time has been spent on verifying and validating AI agent outputs to make sure that it's as good of an experience as possible for the people that will use the tool. Not saying I've designed something perfect, but I definitely wouldn't have been able to create what I've created so quickly a year ago. I firmly believe it's now possible to build amazing things very quickly. One shotting / building prototypes is the modern bullshit factory-- the real cool shit is spinning up hundreds of agents to build something you really care about for months. We're able to make amazing things!! Right now!!! Go create and build!!

Luis Rosias @CrunchBuilds

3 weeks ago

I want to clear up some things about AI coding and software engineering in 2026 that a lot of the smash community seem to misunderstand from the perspective of an ex-Google AI engineer and startup founder because SOMEONE has to bite the bullet and tell you the truth even if you

95 60 894 66K 219

Design tip for Codex/Claude: use blueprints used Sol 6.1 to help me organize my game's village design instead of iterating on the village itself in game, we first iterated on the BLUEPRINT of the design of the village this cuts down iteration time significantly

plus the other billion backlog items / other competing priorities from managers trying to cock measure / layoffs / re-orgs — good luck trying to have committed to something like this in 2022 But now I think high-agency engineers will be able to skip red tape with proof that their contribution is important, since producing that proof is cheap now. so now it just takes someone with enough courage to put the proof on the table in a week, rather than enough stubbornness to wade through corporate bullshit for years

it seems more like we’re able to skip the miles of red tape involved with this type of stuff you’re an engineer at microsoft. you realize the renderer sucks. you want to fix it for everyone. my brother in christ there is so much corporate bullshit that would stand between you

macintog @macintogdev

3 weeks ago

the next year of software will be about proving how much we suffered because nobody competent cared enough to do a good job

57 256 8K 162K 899

@iamvishal16_ios Dictionary that automates spaced repetition flashcards getmiji.app --straight forward product, I found that competitors on the market either (1) were too expensive, (2) didn't have automated flashcards, or (3) were aggressively hard paywalled -- so i made this

Started working on an iOS app about a month ago -- finally submitted it for release! :D

This feels like one of those viral YouTube videos from 2012 that someone would have spent 100 hours editing

Maximus Spritius @MozeTech

4 weeks ago

I just merged two of the highest grossing games ($5 billion+) in history using fable 5.5 (rerouting) in just one shot! I brought Minecraft to the Pokemon world, you can now craft, build and catch pokemon all in the iconic pallet town. There's no longer any reason for anyone not

378 560 10K 1.7M 4K

being curious is more important than being right :D

unironically i think this is possible

Neeraj @neerajjj6785

4 weeks ago

I don't even care anymore

355 622 21K 718K 1K

I FUCKING CARE IN A WORLD WHERE AI CAN BUILD ANYTHING, CARE!!!!! IF YOU DON'T CARE, WHY ARE ARE YOU CREATING??? THE PRIMARY PURPOSE OF ALL OF THIS TO ME IS DISCOVERING WHAT MY SOUL WANTS TO BUILD AND TO FUCKING CARE

This is obviously not as easy as correctness tests; lots of interesting companies working on this problem right now Hypothetically, how do you know you’re even building the right thing? What if you build E2E tests on a garbage system? That’s the point of the new class of tests

There’s a really undertapped market for adding “taste” / “usability” tests Richard’s describing tests around correctness & functionality But with agentic intelligence, you can also detect where confusion might happen, where things aren’t working in the real world quite right

Richard Wu @0xrwu

4 weeks ago

IMO the only tests you should have in the age of coding agents are, in order of priority: 1. Full E2E tests. Nothing mocked out at all. Can even be something that runs in prod on test accounts via Playwright or equivalent. 2. Integration tests. Agents make more mistakes as

190 91 2K 298K 3K