"He's like Godzilla the Hun."

"A human generally can't tell AI generated text from human generated text"

Go read some of the grad student Facebook groups. Folks who regularly see AI text and humon-authored text can tell those apart fairly reliably. TFS talks about how humans agreed with the AI detector about AI-assisted texts being low quality.

When I did editor training, a part of that was to read and edit for *flow*, which gen AI does not currently understand.

> I will admit that I'm getting a bit of an AI-vibe from your post.

Thank you for the personal attack. Are you wishing to end this conversation?

For this workflow, it just needs to be accurate enough to flag a manuscript or reviewer comments for human review. If the authors disclosed and it was AI assisted, great. If not, question what else the authors or reviews might be dishonest about.

The detection AI concurs reasonably with human judgement: "The study also found that submissions in 2025 with abstracts flagged by Pangram were twice as likely to be rejected by journal editors before peer review as were those not flagged by the tool. The desk-rejection rate was higher for manuscripts flagged for AI-generated text in the methods section."

A humon typist or graduate student typing, incorporating edits, and otherwise revising a manuscript learns and improves both themselves and the manuscript by discerning the meanings that were or were not intended with each iteration of writing one paper or across a series of papers, even when using winword's grammar and spelling checkers, Grammarly and similar technical tools. There are AIs that can learn from a user's revisions. Using those would be more helpful than asking AI to generate text and then revising.

The summary also notes that AI detection was higher among papers from countries where English is not a native language. In the previous process where a manuscript by non-English native authors would be sent out to an English language editor as part of the drafting process, the editor would provide helpful questions about meaning, ambiguities, consistency of style, logical flow, etc. AI tools are starting to do that.

For reviews, a human reviewer, native language user or not, will react to unusual spelling and grammar or errors of meaning, and methods/claims that are not plausibly within the discipline, whereas many AIs will parse over all that to infer meaning statistically. AI reviews may also draw connections among concepts that may exist across literatures but do not exist in practice, and/or hallucinate suggested citations about *God cremating the Earth in seven days*, etc.

Corals are also not immune to evolution. The heat tolerant mutants will go on just fine.

One small step towards an iPad being required to restore an iPhone.

What "Decades old C" implementation of AV1 decoding?

AV1 was released in March 2018.

Photographers are already using their software of choice to work with raw formats. Often, that software takes advantage of specific hardware and software features of the $$,$$$ to $$$,$$$ cameras and associated equipment that created the raw files. Staying within the family for software to work with digital negatives and images is not a big jump when someone has already invested in a single platform of camera bodies and lenses, and most photography software that is worth using in the long run provides timely enough support for new camera models. When professional photographers' time is worth $$$ to $,$$$ per hour, the fact that the software or standard is open source might not matter to their teams' workflows or bottom lines.

I've installed and supported GIMP for dozens of groups in the past, but I've stopped doing that because kids who learn that software gain significantly less advantage for employment and further education compared to learning more popular desktop and cloud based image editors.

Programming and the required design skills feels like an art to me, so I would advocate for theory and practice being taught together in the same way that learning music theory doesn't make a person capable of playing a piano. But I may be an exception.

Around the time of the last *dinosaurs mating*, the debate was whether churning out compsci grads who knew how to use C / C++ was sufficient, or if they needed to learn how to program in C / C++ and other languages. Some folks thought that the compiler-writing course should no longer be mandatory because *no one needs to know assembly anymore in the real world.*

I wonder if the next debate will be about whether it's necessary for students to learn how to read documentation, or whether knowing how to ask gen AI a question or use code completion is sufficient.

> It won't. People run Windows 7 do it for one of those reasons:

- the machine and its particular version of software are part of a bigger validated system or configuration which has an expensive / difficult certification for the industry, utility, health, marine, aerospace, etc. environment.

WINE will not replace Windows in such environments any more than SD cards will replace floppy disk drives on aircraft which are already flying.

GPTs are a logical step combining spell check, grammar check, autocompletion, search engines, writing notebooks, word processors, and encyclopedias, among other scholarly tools. They can help to autocomplete *some* ideas that one already has, but the user still needs to already know information to create a useful prompt, and how to adapt the resulting text into the real world.

Remember amd.slashdot.org?

And my decompression sickness.

There can be value in investing in domestic infrastructure. Tariffs with other trade and non-trade barriers can encourage that investment.

Recall that folks are discussing tariffs on Chinese imports *because* China started to figure out supply chain resiliency a couple decades ago in ways that made the rest of the world uncompetitive in green energy and related manufacturing.