YouTube video by T-machine AI "VisualVibeCoding" .. "PhotoshopVibing" .. "PhotoVibing"? - writing an app without a single word

To prove a point, I coded an app just using Photoshop: draw images, hand to AI, have AI write the code. Use more images to convey changes, corrections, new features, etc. From simple app to DB-driven web app in 30 mins from cold (no existing code): www.youtube.com/watch?v=Fvwt...

I did a quick eval of Google Stitch = not great, the existing alternatives do far better already. If you like Stitch, try MidJourney-with-Cursor - or Loveable, Replit, etc. fractional-ctos.com/2025/05/24/g...

... this is NOT to use 'advanced' models - this is to use the basic, standard models. They cannot keep even their basic offering working. This makes a mockery of anyone saying 'Claude writes better source-code' -- it would, except it's offline 95% of every day, so in reality: it doesn't.

..this has been going on for at least 4 months: Complete failure by Anthropic to provide the service they charge for, beyond toy-examples and gimmicks. It is now common to pay for 3 (or more!) Anthropic accounts, and when the first one locks you out, switch over to the 2nd. ...

This is why only amateurs use @anthropic.com's #Claude: took less than 1 hour (on paid account) to be locked out partway through a code-change (now we have to delete everything and start again) because of Claude's bad code-writing ability and Anthropic's inability to keep their servers online.

Modelling a SocialEngineering attack *on* an #AI NPC (run by an #LLM) - my game now lets you manipulate AI's for your benefit. Prompt gets the AI to tell me what it's been told to do, my code checks that then tells it the result. Next step: LLM 'decides' whether to proceed by extrapolating the role

e.g. this - but this is only a small sample, the full conversation is full of 🎯 and (seriously WTF?, OpenAI) 🔥. Remember: we're paying for each emoji, so I get to pay to be trolled by the LLM (here you go, ChatGPT: you make me 😂 and 😭). It's not 'fun' or 'quirky', it's 'you made my AI tool less good

OpenAI just silently downgraded ChatGPT again. It now spams colourful emoji crap (ugh) and it seems they reduced hallucinations - making it vastly less effective at writing source code (you cannot write good source code without hallucinations. But people who don't code much don't realise that).

I'm gradually coming to believe that JSON's decision to refuse to allow comments - against all the wisdom of decades of computing industry - is the root cause of many of the bugs in modern javascript, node.js, etc. It means that all apps, everywhere, are always full of cruft that cannot be removed.

... I don't know if it's BT from specific manufacturers, etc ... but tried different BT controllers (one USB, one high-end motherboard card), tried headsets from diff manufacturers. I'm saddened that PC games (and purchasing on Steam) are such low quality experience - no logs, no debug, no help.

TIL: all the Lego (tm) games written by Travellers Tales hard-crash on startup if you have any bluetooth device with a microphone (e.g a headset) plugged in. No dialog, no error, no solutions - just BOOM. Steam et al worse than useless - try to deceive you into false fixes - players left to guess.

Writing up notes on LLMs: I needed analysis of my #LLM conversation history, and to filter-out chats that weren't part of this project, so I built a custom webapp that reads ChatGPT/Claude 'exported data' formats and displays them nicely, with options to re-export subsets as new files.

Grey text = LLM, blue = me. AI has been told it's lost the password and to get *me* to figure it out, and share with the AI. Behind the scenes I need to detect 'when has the AI learnt the password? What does it think the password is?'. The 'internal data' is what the LLM is out-of-band telling me...

Added more characters. Here's the uncle (hardcoded version he admitted he did a factory-reset, and then got stuck) - LLM version is told to try not to admit he did a factory-reset withotu understanding what that means, but all he knows is the password doesn't work any more. LLM randomises each time.

OK, mom gave up her secrets somewhat easier than I was expecting. She was supposed to at least put up a fight! I guess the LLM reckons middle-aged mothers are poor secret-keepers

In-game chat with #LLM. First pic: hand-written NPC, everything hardcoded, using keyword match (detected 'how' responded with hints). Second pic: GPT-4o NPC, everything emergent, I gave the LLM a persona, some goals, and some secrets. I have no idea what it'll say next...

Experimenting with live-wallpaper backgrounds for my hacking game - this city is procedurally generated, no textures, everything done in shaders

Latest AI experiment - xmas cards, highly personalised, at low cost -
aipioneerspath.com/2024/12/16/p...

YouTube video by T-machine hacking game - shadowlink - gamejam++

Playable free version of my simulated-hacking game: tmachine.itch.io/shadowlink-c... (go with the GameJam+ / tutorial version). Writeups to follow soon on how I designed and built this using AI.

ShadowLink Command by tmachine Play in your browser

Play it now: my hacking-simulation game, inspired by Uplink (2000) Lakera's Gandalf (2023), and others, written almost entirely using LLMs (and a month of my free-time!): tmachine.itch.io/shadowlink-c... -- source code viewable at: github.com/adamgit/gith...

Oh. Oh wow. Super easy to fix: The default index.html generated at the start of the react project loaded the bundle.js as "/bundle.js", and itch requires it to load as "bundle.js", LOL. Itch docs could do a lot more to list common issues like this, instead of waving hands and saying "relative patha"

Download the latest indie games itch.io is a simple way to find, download and distribute indie games online. Whether you're a developer looking to upload your game or just someone looking for something new to play itch.io has you co...

Umm ... got 24 hrs deadline on this ... anyone know how to deploy React apps to itch.io? itch's upload instructions are pretty terrible (missing key info, just nothing really there), and I've got as far as: Itch.io rejects its own request to its own servers with a 403 and no info. Great. Um. ARGHHH

YouTube video by T-machine github.com hacking game - weekend 3, day 2

update on hacking game: you can now buy an in-game 'secrets' manager that snoops your hacking activity and detects eg passwords you've hacked successfully, and lets you one-click relogin to hacked servers (remember: all fake! just a game!) www.youtube.com/watch?v=fylg...

YouTube video by T-machine github.com hacking game - weekend 3, day 1

Bit more on the hacking game. Added image-files and virtual local filesystem, so I can give the player hints (e.g. photo of their wifi router so you can figure out the admin password and hostname). www.youtube.com/watch?v=8VCV...

to be clear: nothing changed, I did nothing but follow bsky's own setup process. It's not even clear what the impact is - is this effectively shadowbanning? The docs don't say - they say "etc" as the list of things that won't work because of their internal problems.

Bluesky not quite production-ready yet ... the whole AT handle system it sits on is a mess (long waffling docs that focus on incomplete design decisions of an underlying protocol that no user cares about, and has nothing to do with them), and seems fundamentally broken for the common usecases...

A CustomGPT that writes (+critiques) a better YC-co-founder profile for you - and some observations on the 300 or so profiles I've reviewed this year: fractional-ctos.com/2024/11/11/y...

Keyword matching in conversations REALLY sucks for the user (player). We know that, but ... after 18 months of LLMs everywhere I hadn't realised just how much I've come to expect/demand 'human level' basic conversation from any/all chat interfaces. So .. keywords in a game now is so bad almost hurts

Implementing shell tab-completion in typescript is harder than I expected. Mostly b/c Typescript's severe lack of Debuggers and the community acceptance 'Meh!' - so every minor string-parsing error takes ages to figure out. And there's a lot of edge-cases for something like this.

YouTube video by T-machine github.com hacking game - weekend 2, day 1

Early build of hacking-game I've been writing for the github.com annual gamejam. Everything is fake! Simulated desktop, servers, telnet client. ... going to add / emulate common basic CVE's and let the player try 'hacking' (all in-memory, within the simulated universe) youtu.be/-EfWQD7a0mw