Depends how much it matters. See if you can get the business to put it in dollar value terms or at least as opportunity costs against other eng work.
Would be sort of a spike project, but would be interested to see how well off-the-shelf VLMs could handle screenshot comparisons, and then a bit extra to see how to deploy them efficiently and effectively.
way less work
Okay so do postgres tsvector then
Still an improvement
You probably still have other things you want to do, too
Also, the more advanced the feature is, reality is that the fewer users can make use of its full capability
But forget all that, if you feel like doing it anyway, then do it, this is FOSS
Refactor first
Engineering has to draw the line somewhere. Excuse my analogy, but if the order is "you are not allowed to clean up the construction site, only build", and there's so much crap around that it's no longer safe, then the answer is, "no, I can't work like this, find someone else to do it"
That said, there is an art to working in and managing tech debt. The amazing thing about software is how flexible and how twisted it can get behind the external interface
the GUIs people build aren’t good enough
Big agree
TUIs are inherently better than GUIs
Perhaps not inherently, but TUI users have much less tolerance for "lag", so TUI apps are built to that expectation. That's often not the case for GUIs.
Didn't watch whole demo, but one thing: rotational gyroscopic torque
He showed Box3D correctly modelling: https://www.youtube.com/watch?v=1VPfZ_XzisU
And compared it to a game in Unreal when a spinning rifle had an unnatural perpetual motion spinning quality to it
Check out out: https://box2d.org/posts/2026/06/announcing-box3d/
Why is right-to-left for WHOLE PARAGRAPHS possible for Arabic languages, but not for CJK nor the Latin alphabet?
"Hi, agent. Launch two additional agent instances. Prompt them with this message."
Absolutely silly premise. Continuing the author's analogy, how would pacing oneself in one marathon improve performance in the race tomorrow? What about next week (a race of indeterminate length)? Or next month (turns out will be a triathlon, but you don't know it yet)?
It doesn't, and the "races" that engineering teams run are more like those latter examples than they are like performance during a single marathon.
A solar panel that occasionally blocks a window would be completely useless the other part of the time.
A window permanently blocked by a solar panel is better known as a wall with a solar panel on it.
A vertical wall with a solar panel will not be as effective as one angled towards the sun, such as one on a roof might be.
It's like Twitter, but anybody is free to spin up their own instance, and they're all inter-compatible, so any one can view content from any other.
That freedom and openness gives users choice. If one instance owner turns out to be a real dick, the community is able to move off of and blacklist it.
Email is an oft used example. Leaving Gmail doesn't lock you out of email because you can switch to a different provider or (with enough effort) even run your own email.
"$0" just means they don't charge a flat fee on top. You'd have to check the conversion rate they give you compared to market rate to really know.
It can be the case, even likely, that your card's rate is better than local currency converters and atms targeting tourists, but point is they're still not doing for free.
In the short term, a small amount of duplication to prevent bad abstraction spaghetti is worth it
In the medium term, true duplications are better extracted after the fact rather than correctly guessing future reusability.
In the long term, total extraction into separate vertical or external package, as suggested by the author.
Over time, this progression is natural and unforced. If you don't need an abstraction to be de-deuplicated, then don't. If you don't need to promote an abstraction into its own vertical/package, then don't.