New month, new app. I got annoyed with media players. QuickTime for its lack of file support. IINA for its lack of Timecode, JKL playback, and trim tool.
Aagedal Media Player (open source) combines the best of both, while adding even more file support, and being faster to launch. player.aagedal.me
I’m not really surprised, but the recent confirmations about the AI use in the military does give me some renewed unease. Skynet vibes.
It’s like humanity prepares to make the post-human transition be as fast as possible. We are not at Skynet level AGI yet. But maybe within 24 months?
Not sure what makes me cry more: That one scene in Interstellar, or the season finales of Doctor Who.
The One Ring is somewhat like a horcrux, right? Was it more powerful because only one was made? Should Voldemort have forged the artifacts himself from magical magma, or was it better to use the older magical artifacts? Could Harry have dropped the horcruxes into magma?
Since journalism is (ideally) about promoting truth, and about being a sort of check against power abuse, there should likely be more journalism about the potential issues regarding a hypothetical AI truth monopoly. A modern version of 1984 doesn’t seem too far off.
While AI have and will make the lives of real journalists easier, I can’t help but fear that in the near future the best "expert" a journalist can ask is an AI model. Human experts may also be influenced by AI and repeat its output. Those who creates AI (will) have a lot of power over truth.
I often wonder if journalism fails important parts of its mission, because news organizations lack the surplus to compensate for the exponential information growth. Being unable to see the pattern.
Or was the nature of journalism not designed to handle this amount of information processing.
Lots of things are important, but it is hard to say what the most important problems are when everyone has FOMO on the latest news. "Tribalism" also doesn’t seem to be helping.
I’ve previously called the internet economy a war of attention. But exploited more systemically across platforms you have powerful tools of distraction and misdirection. Overload human minds and its capacity to care about important things.
Finally got around to making a simple website for my each of my macOS apps. mediaconverter.aagedal.me photoagent.aagedal.me imageextractor.aagedal.me
I recently learned that Pandora’s “box” is a mistranslation of the original Greek “jar.” Pandora’s jar was opened, and the only thing supposedly still trapped inside is hope.
Not sure if there is an AI virus going around, but both GPT 5.3 Codex high and Opus 4.6 did really stupid things today. Yesterday I was impressed with both, but now I’m missing 4.5. Hopefully this is temporary.
”GPT‑5.3‑Codex is our first model that was instrumental in creating itself. The Codex team used early versions to debug its own training, manage its own deployment“ … ”our team was blown away by how much Codex was able to accelerate its own development.“
openai.com/index/introd...
Future ideas (not sure if practical/possible yet): extracting image description from audio sidecar (AI transcription), C2PA signing (bring your own certificate), similarity grouping/sorting, basic editing (with XMP edit metadata) and automatic contrast normalization, image analysis tools.
It supports metadata templates, metadata variables, reverse geocoding, face detection and clustering, FTP upload, star ratings, labels, full screen preview in HDR, C2PA detection (with mechanisms to prevent accidental invalidation).
New year, new app. Aagedal Photo Agent is a tool to quickly add IPTC compatible metadata to files. A free and open source alternative to Photo Mechanic or Adobe Bridge. Though, only for macOS. github.com/aagedal/Aage...
Tests were done in DaVinci Resolve in a 4K timeline. Since I’ll likely never export any 8K films. But downscaled 8K can often look better than native 4K.
I did do a brief test in a HD timeline too. That basically removed all benefits of 8K compared to the 4K formats.
Note that these tests were done using the lowest bitrate setting available for 8K 4:2:2 25p. And using regular ProRes RAW, not ProRes RAW HQ.
What surprised me the most was macro blocking visible in the noise in ProRes RAW. This may explain why there isn’t a LT version. (XAVC HS had ≈no noise.)
… the 4.3K ProRes RAW preserved a lot more natural noise. Making it look more like film out of the camera, with no processing.
That said, when using the Film Look Creator in resolve to add fine grain and slightly soften the unnaturally sharp image, the 8K ended up looking more cinematic in the end.
This applies to both the 4K and the 8K XAVC HS recordings, but the 8K file was even better than the 4K.
Comparing 4K XAVC HS to 4.3K ProRes RAW, I strongly preferred the RAW image. But the story was less clear when compared to the 8K image. Both had benefits. 8K preserved more real detail, but…
… that the professional XAVC-I format graded worse (a lot more visible macro blocking) than the "consumer" XAVC HS format. Both were 10-bit 4:2:2.
This means that unless you record RAW, Alpha-series cameras recording XAVC HS will be able to deliver a better image than most of the FX series.
Just compared ProRes RAW from Sony FX6 and A1, along with XAVC-I (h.264 based, in FX6, 4K) vs XAVC HS (h.265 based, in A1, both 4K and 8K).
ProRes RAW looked very similar between cameras, perhaps with a slight edge to the A1. But basically indistinguishable.
More surprising however was …
But while that task you made an app for has become easier, you have likely ended up with another app you need to open. More apps could mean more time used to managing apps.
And when cross-app management itself can be automated, you may not be needed anymore.
How easy is it to automate all you do?
One year ago I interpreted CC as Creative Cloud. Now CC is Claude Code.
Vibe coding is taking over the world. Software on demand is (to a degree) possible. A simple app can hypothetically be made in less than a minute.
IQ tests already suggests the best AI models are smarter than most humans.
Luckily(?) they are still struggling a bit with with long term memory and real time task adaptation.
And with the introduction of AI doing people’s work, and potentially fixing everything tech related. (Even professionals are turning to vibe-coding now.) People are likely to understand the technology less too.
Why think, when computers think better?
There may be people on earth that has never seen a real high quality image. That scares me a bit. While media literacy in some sense has never been better, in other ways it may have become worse. Even without better AI images, just the amount may make people forget what pictures used to be.
On top of this we also have web image compression, which itself can make surfaces look unnatural smooth and edges unnaturally sharp. The smaller the image is compressed, the more fake it looks. Invisible automated upscaling may also make people think that the image had details it didn’t.
AI images often have unnaturally smooth surfaces and sharp edges. The same goes for computational photography. Natural edges aren’t that smooth or hard, but on a small screens or at low resolution it makes the images look sharper. (Sharpness is not detail.)
AI images often lack hard shadows, because they are often bad at making physically accurate shadow directions. What does computational photography do? It blends together so many exposures that almost all shadows disappear. To preserve the most dynamic range.
