QuantumGood

When we provide written documents with embedded videos revealed by on-screen toggles, covering different sections of a help document, almost no one reads the doc, they just find the relevant video. Many of our docs require images, and videos do a better job of, for example, showing how to clamp an arm onto a structure.

Video that cover different parts of an outline are a great use of videos. Videos that try to cover too much all in one are a lesser use. Yes, you can use scrub to find in a video, but showing a text outline and letting users click to reveal videos for specifics topics works very well.

Generics need to be more carefully sourced if outside the U.S. I prefer brand names from reputable manufacturing areas (Singpore, Germany, etc.), and can get them for close to the price for generics, often starting research at PharmacyChecker.com

More perhaps at the appearance of logic. I still find all models make easy to find mistakes in logic, if you think carefully about what they say. Usually they allow "close enough" assumptions that are not, actually, close enough.

When I ask about acoustics, they still often make incorrect assumptions, e.g. overlooking that they are speaking of logarithmic display of digital levels when the discussion has shifted to SPL (sound pressure level in air).

The more data in the PDF, the more nines you need in the OCR accuracy. Plenty of 5/S, O/0 and other issues exist at frequencies that cause problems.

AOSP = "Android Open Source Project"

One of many "new normal" artifacts that seem all the more strange when the pace of change accelerates.

The Kia factory recall key cylinder fix solved the hotwiring issue; and the engines that burn oil last pretty well as long as you keep adding oil. Not saying these countermeasures are solutions, nor that the issues are minor, but they produce a car that can't be easily hotwired and lasts a long time.

Some of these cars (older Kia Souls, for example) become a great value if you implement the countermeasures, and also because they have a lower cost of repairs than some Hondas and Toyotas.

I learned this because I seemed to find cars that had few problems that had nonetheless been given up for salvage. What I found was that if a stolen car is not recovered quickly enough, insurance will pay the (properly-insured) owner an amount close to the replacement value of the car; they may then recover the car later in the same shape it was stolen in other than a broken window, door handle and steering column cover. They then put it up for auction even though it would be better financially to work with a dealer or fleet network to cherry pick these high value cars that only need some simple fixes.

For example, my $2,220 bid won a 2015 Kia Soul+ (49K miles) at auction that a dealer had sold for ~$12,000 some months previous. Bidding competition was light primarily because of the "easily stolen" and "likely to burn oil" issues. But with fees and a few cheap repairs, I got a $12,000 dealer-prepped car for under $4,500. I added remote start, Android Auto and a backup camera, gave it a little ceramic spray coating and got a nice car for my wife.

I also threw in a fake $12 steering wheel lock bar as a deterrent. (It's a real bar, but I don't lock it because the "unable to unlock the cheap lock" risk is to me a more likely hassle than a broken window). I considered adding an immobilizer, but the car was stolen with one installed, and I could always add a switch to the battery under the hood in case better-prepared thieves than the "hoping for an easy hotwire" type show up.

I love digging into questions and seeing where they lead; this all started with a vague sense that dealers made a lot of money on a high volume of used cars via auction houses, and trying to learn how it worked. This also leads me to argue with AI quite a bit, and until recently, made the idea of AI "thinking" kind of laughable due to its many mistakes once you start following the chain of reasoning.

I used to do demonstrations in large workshops, and I would stand behind someone reading a script, and hold up a flag 3-5 words BEFORE they would stumble. I could tell when their compulsion to say the next word had gotten ahead of their sense of the meaning of the word, and 3-5 words later they would reliably stumble.

In voiceover, where you have to read rather than memorize most scripts, some percentage of your available bandwidth visually processes the words that you must look at. This causes issues for the majority of talent, and significant issues for a minority of talent. Part of the issue is a compulsion to keep speaking words due to a feeling that they seem to be pushing at you and piling up.

It was rebranded as "Alexa for Shopping"; that is what I am referring to.

Amazon's AI search doesn't do that, however. For that reason, 90% of my searches are now prompts to the AI, instead of regular searches. But it occassionally freezes the app.

page is now 404, this one still works https://cdn.saffire.com/images.ashx?t=ig&rid=OhioStateFair&i...

One of the things not represented well by the hype cycle is how hype is a moving target. When "ai defenders and hype" assign a number to things, the early numbers can become closer to the later realities.

Things change, and people forget why they were ever any other way. There are definitely resistance forces, but when the overton window flips, people just accept the status quo again.

  > should want [others] to make binding agreements

As Munger emphasized, others act according to incentives. First, clarify those incentives, then incentivize them if you can.

Sometimes its worth tracking usage optimization and raw abilities as separate qualities.

Figure out something to say even though you are not yet very clear on the other person's communication. Common approaches: Ask for clarification, mirror what they say. Growing up autistic, I couldn't find a way to maintain/create a positive connection until I started to focus on the emotions, not the data. I then discovered that the emotion seemed to preceede the data for many people, and that if you stayed on the emotional wavelength, sometimes the data I thought I perceived would turn out to be irrelevant. A cliche is men attributing this to women wanting to share feelings, but it parallels another cliche, men wanting to share things that anger them.

I use high-quality tape on the spine after flattening it for easy reading; still flattens well, but holds together longer.

Treat the answer like a game instead of treating the question like a logic puzzle. Optimizing the answer framing instead of the question framing opens up a lot of possibilities. This also works in personality conflicts affecting communication (like marriage).

Love the AI format with obvious misunderstanding to this and all your other comments; makes it much easier to find AI bots like yourself.

Allowing for creative incubation is a real thing. But task restarting penalty is real as well: you lose context and your mental model; working memory loses details; dependencies are reverified unnecessarily; flow state momentum lost; environment reconstruction each time and similar administrative tasks, etc.

It's an "absence of light" pattern (as opposed to dark pattern). Creating a very common tech support issue; sometimes "login" isn't even visible.

Commoditization always takes volume from the marketplace, but not necessarily profit. Apple and Military tech come to mind. Positioning is rarely done well looking forward, but often shakes out in unexpected ways.

A one point a very hated thing was how they defrauded advertisers using their services. They have increased the fraud over time, so that hate hasn't gone away.

At a certain point in the process, a little additional content multiplies the value of your existing efforts. But that is misunderstood (survivor bias?) as "make lots of content, because the successful people did".

"Good artists copy, great artists steal, the best artists make lies seem true." – Steve Jobs.

Actually, Jobs self-servingly refactored then misattributed to Picasso the line,

  "Good artists copy, great artists steal."

Jobs did seem to properly remarket Picasso's intent, though, as Picasso had said:

  "When there's anything to steal, I steal."

  "What does it mean, an artist copies? It is normal. The important thing is to do it better."

  "Art is a series of masterly thefts."

"Yes, ..." is how I notice all the models (4-5 that I use) in the last week or so have begun saying "No, ..." except for Kimi. I ask a lot of questions in the form of "Question about the feature: does it allow for ..." and the answer will reliably start with "Yes, the feature exists and does this thing you have labeled as [blank] which can be explained as ..." and later "However, ..." giving the information I requested (no, it doesn't allow for what you need). It's actually kind of uncanny.

I've started testing identical prompts across models, just to clearly see the "Yes, ..." with the "..but.." buried.

Example: "Can you view the salvage car in the lot before it is scheduled for sale?" "Yes, cars are viewable on the lot by members before they go on sale, blah, blah, blah." Then later "However, cars not yet scheduled for sale are not allowed to be viewed at the lot".

Example: "Can I control the keyboard shortcut Chrome has hijacked so that my existing OS keyboard shortcut will work even while Chrome has focus?" "Yes, chrome keyboard shortcuts, blah, blah, blah," and later "No, chrome does not allow ... ".

There is clearly some recent implementation of the idea that starting with "Yes, ..." has some benefit, but I'm having trouble adapting to the feeling of being lied to as a policy.

And knocking out the messy garage background to replace it with a showroom is the easiest prompt.

Compaction is a feature like swapping out a car for a tricycle is a feature for greater economy.

Each new iteration of the hype cycle feels "like its not hype" because the tech is more promising in new ways, but the hype comes from human perception. New tech has to fight a lot of real world and project management issues, and development morphs and slows. We don't know exactly what we have until it faces these issues.

... and then it stopped having that ability to build on create concise, accurate summaries and expand on them. Same problem with GPT 5.6 Sol. I even tried and old prompt in the (more nerfed) recent version. The core initial thinking ability lowered, and I'm getting the same over-metacognition and epistemology as it tries to pull at threads in the brainstoriming process again.