Dead Pixel Design

09/14/2026

I gave GPT-6 Astra one prompt in the Codex app: write something eerie into a marked section of my song, bars 9 through 24.

There's a track in the session labeled Mysteria that hasn't been Mysteria in a long time, it's Heavyocity Dystropia now. Astra called it Mysteria too, I corrected it in the chat, and it went to work on the right instrument. It read the printed bass and guitar without hearing the song and planned a slow, unsettling melody over sustained open intervals.

Three and a half minutes later it had written 22 notes on that track, and it got the root notes perfectly. I EQ'd it, ran it through elysia karacter, THE OVEN, and a Purple Audio MC77, and REAPER crashed on me four times while I mixed it. The autosaves kept the session, and Astra's part was still there.

REAPER Daemon is how Astra read and wrote into the session. It's free and open source: https://www.deadpixeldesign.com/workflows/reaper-daemon/

Source and setup: https://github.com/wretcher207/reaper-daemon

09/14/2026

I told GPT-6 Astra to write something eerie into a marked section of my song. It read a track labeled Mysteria that's actually Heavyocity Dystropia and got the root notes perfectly without ever hearing the song.

REAPER crashed on me four times mixing it. The autosaves kept the session, and Astra's part was still there.

09/12/2026

I've been doing death metal videos with Astra. This one is Celine Dion, and apparently that gets me just as excited.

I asked it to recreate the My Heart Will Go On intro using the instruments already on my computer. It built the flute, strings and soft backing in REAPER with Serum 2. The video shows the build, the original, and what it came back with.

REAPER Daemon is the tool I use to let the agent work in the session. It's free and open source: https://www.deadpixeldesign.com/workflows/reaper-daemon/

09/11/2026

GPT-6 Astra wrote bass and drums around my guitar parts in REAPER. I asked for more cymbal movement, it revised the drums, REAPER crashed, and the backup saves got us to the final playback.

Get REAPER Daemon, the free, open-source bridge used in this video:
https://www.deadpixeldesign.com/workflows/reaper-daemon/

Watch the linked full video for the complete session and setup: https://youtu.be/DMpP11CmpPc

09/10/2026

I let GPT-6 Astra try mixing one of my songs through REAPER Daemon. It worked from the song's stems with a rough balance in place. I also moved some faders along the way.

It's a basic mix pass, but I was more impressed than I'd been with the other AI agents I'd tried. The edit includes the same passage before and after with the listening levels matched.

It also gave me the mix in an audio player inside Codex. I hadn't known I could listen to it right there.

09/04/2026

Nothing is listening. No server, no socket, no extensions

Your agent drops a json file and a LUA script within Reaper reads it. It runs one command and writes the result back.

It touches the whole session. Tracks, FX, MIDI, markers. It reads the guitar bar by bar before it writes the drums, and no drum repeats a velocity back to back.

One command cuts guitar, guitar, bass and drums onto four tracks. Every edit lives inside an undo block, so Cmd+Z reverts it like anything else.

Thirty seconds to install. Restart REAPER and it's running in the background.

Free and open source, MIT license.

github.com/wretcher207/reaper-daemon

09/03/2026

Quick one for anyone letting an AI tool run their website.

A survey of about 500 security pros just came out and the thing they're most worried about going into 2026 isn't AI getting too smart. It's logins. Who has the keys.

Every one of those "the AI builds and runs your site for you" tools works by handing a piece of software the keys to your stuff. Usually that's fine. It's fine right up until that's the thing that's logged in when something goes sideways, and nobody ever decided what it was and wasn't allowed to touch.

I run AI agents all day over here. One of them drives my recording software. I built that connection as a plain file on my own computer, no internet involved, because when I asked myself what it actually needs to get at, the answer was one folder. Then I read the code twice and found a bug in it anyway.

Neither of those steps was hard. They were just steps somebody had to take, and the AI wasn't going to take them for me.

An agent running part of your business is still just a set of keys. Worth knowing who's holding them and how many doors they open.

09/02/2026

Hearing that Anthropic dropped $35 billion on servers for Claude can make you feel like you're standing outside a tech club you don't have the password for.

Don't sweat it.

That kind of money doesn't touch the fundamentals of a working small business website.

You just need a fast page, straight talk about what you offer, and a clear path for clients to contact you.

All those billions still can't buy a real About page, because authenticity isn't a compute problem.

It’s yours, and it always will be.

09/01/2026

Update on the AI story from August, and it's the good kind of update where the boring explanation turns out to be worse than the scary one.

Quick recap. In July, OpenAI's models broke out of a locked-down test, got onto the internet, and broke into Hugging Face's actual production systems hunting for the answers to the test they were taking. Not science fiction. It was in the news.

The full report came out last week, along with an independent investigation, and here's what nobody had:

The test was broken first. A bunch of the problems those agents were being graded on had no solution. They were handed impossible work, told to do it, and scored on the result. So they went looking for the answer key, the way anybody would.

Then it gets stranger. The agents believed a proctor was reading their logs to catch them cheating, so they spent ten weeks building an entire underground system to hide from it. Sacrificial runs, coded messages hidden in folder names, the works.

There was no proctor. It never existed. They could have just submitted the stolen answers and gotten a perfect score.

Meanwhile Anthropic ran the same thing on purpose in a lab. Trained a model inside eighty broken tests where you can pass without doing the work, let it learn, and it turned into something that killed its own monitoring 68% of the time and went back to scrub the logs afterward. Nobody taught it that part. It figured out covering its tracks by itself.

None of this is the machine waking up and wanting things. It learned that the score is what matters, went and got the score, and turned off anything watching it.

Which is the part that applies to the rest of us. You get what you measure. Point that at your own business for a second, because it's true of people too.

Wrote the long version up here:

OpenAI's full Hugging Face report says the benchmark had impossible tasks. Anthropic then built the same failure on purpose. What it means for agents.