Andrew - EOSHD

I am sure your username rings a bell? Were you around in the older EOSHD days? How did you get into the R6? Via Magic Lantern?

Panasonic, the mad bastards - have 4K modes in the GM5 firmware, just not linked to any menu options. On the Lumix S9 meanwhile, finding all sorts of extra video modes that they're not using!

Just taking a deeper look at the Panasonic GM5 firmware, such as great little camera. Going to find out if it can be modded to shoot 4K.

Agreed. But the solution is quite straight forward - ask Claude to study the architecture and refactor occasionally to keep everything on track. The limited context window problem will go away one day. It'll have a proper long term memory. When you're human there's an understandable desire to be efficient. With my F1 game for example I have a custom 3D engine in C++ that powers about 3 other racing sims as well. I thought about making the engine like an API - one thing that serves 4 different games and you only need to change one underlying codebase once to add a new feature or fix that benefits everything using it. But even this approach might be worth reconsidering when it comes to AI. I was going to make the engine stand-alone like Godot but then I thought differently of it, because the cost to maintaining 4 different engines is now much less than if I were hand coding it all. Sometimes it's fine to just let it diverge and fix stuff as and when it comes, even if you have to tell it to look at similar code in 4 different places, it can all be sorted in 30 minutes.

Computing was once about a machine always waiting for the inputs from a human, in a deterministic way, it sits and waits - you enter the inputs. Coding was like this, you write all the code, you look stuff up on stack overflow and cut and paste algorithms occasionally and then with a lot of testing and iteration you get what you want after many many months. For those whose skill was 90% this and 10% ideas its obvious they'll want to carry on the old way and that they're on the defensive about the new way. For those with 10% coding skills and 90% ideas it's working out much better. 1. You can't say "AI gives a false sense of having super powers" as a blanket statement 🙂this applies only to those who lack technical thinking, ideas, taste, purpose and so on. As a blanket statement, it falls down. It's also very easy to get out of date with your experience of the tools - they can evolve in extreme ways in a matter of weeks, as was case of Claude Opus 5.0 and 5.5, the step to 5.5 was a massive step change so all the people basing their judgement of AI's capabilities on 5.0 were out of date over night. 2. I agree it removes the need to "deeply understand how systems work and click together" but only in the details, not the broader picture. It is better at understanding how systems work and click together than most humans on this planet, at this point. What you need now is the type of brain that can take the new way and run with it rather than a mind that gets bogged down in the details or tries to duplicate what the AI is better at. Coming from a position of ideas and imagination is great, because you don't feel on the defensive about your syntax skills and whether its better than you at your main job or not... Ideas people don't feel threatened by it and are happier to let Claude get down to business without constantly micro managing and looking over its shoulder... or worse, looking at the code. We have to face facts. Computing is never going back to human comprehension, writing or reading of millions of lines of code, our minds are no longer the best tool for this. The capability of the AI to understand more complex problems than a professional programmer or even a whole team of them means you can be much more ambitious. Between February and October in one year, I went from making a teleprompter app for the iPad, to porting Quake to the S9 and developing an NLE, and an F1 sim. My coding skills didn't change, they're still in the pre-2026 days, but my tool use did. It has lifted me up from the muddy field and put me on the road at 200mph to finished products. There are some like the lead dev of Magic Lantern that simply refuse to take the wins, but everyone is entitled to their own principals and beliefs. I'd just prefer to be on that 200mph road to amazing software, than to still be stuck in the mud figuring out how to get a 1100D to write a .txt file to an SD card.

I just need a "TONEH!" sound FX when pressing the shutter button. https://www.youtube.com/watch?v=aQ8VodC19-g

I won't be releasing it until: At least $100,000 in donations And at least 5 credits from YouTubers amounting to 1m views.

"But why?!" God it annoys me. Has the general knowledge got that decrepit that they don't understand what a proof of concept is any more? It also pisses me off that he uses my clip and doesn't credit, but I expect nothing less from the click weasels in 2026.

I had the same thoughts when I started using Claude and then a follow-up thought that wasn't quite as nice!!

I tried Deepseek Flash 4.1, and it was alright... You can't take a janky knife into a gun fight though... The camera's life depends on it. Still a massive gap to the frontier US models.

Here's the recipe, I'm going to lay it out in totality for the click weasel snapgrammers who definitely always credit their sources. 1. Download Claude and stick it up your fucking arse. 2. Buy an S9, your first Lumix camera 3. (The real secret was Le Chonk, the French AI model) 4. Sit back and let the youtube ad revenue flow in

I misread that as cripple hammer and chisel

Does that fit in the hotshoe?

Connect it to HDMI and bingo you have a games console

Pretty much anything you can dream of and can write in C code or Arm64. It even has a bluetooth host and USB HID capability which means all sorts of accessories can be accommodated with custom functionality. Steam Controller 2 as a follow focus anyone? 🙂