My takeaway: get people to independently create an artifact representing their priorities, share the results, discuss the differences, come to a consensus informed by everyone's insights.
Whatever artifact makes sense. Maybe a spreadsheet list. Maybe a one page, or half page, serious coherent characterization from each person.
How people independently see priorities before collective discussion, being as important as how they rate them, I would go with one page thoughtful summaries.
Having experienced partially-aligned organizations, I would want key contributors to a discussion like this to come from every major part of an organization. Emphasis on very different roles, especially important leaf roles, not just top leaders. Otherwise you get high bubble-at-the-top alignment. Top-only strong alignment is like concrete, perniciously inflexible. Which is what I experienced. End-to-end insight and discovery of alignment is needed to get adaptable, in-touch, actually effective alignment.
And as is hammered so well in "Creativity, Inc" by Ed Catmull, alignment isn't just about coordinating efforts. It is about continuously identifying and removing friction, of any kind, anywhere, for anyone, as individuals make their very different contributions in service to the common direction.
> get people to independently create an artifact representing their priorities
Except trying to get most stakeholders to be alone with their thoughts to be condensed upon a blank document before them causes them to violently open outlook and schedule a meeting with everyone instead.
Pre-alignment-meeting alignment! A good thing to explicitly exclude.
Definitely need a healthy culture when asking for honesty and openness about what people think, including they see as unclear, unknown, inconsistent with others, and/or outside the existing box.
I am going to second the comment you are replying to. Strongly.
Why are you (indirectly by omission) asking a cohort of people who need information to be direct, to redirect? That's a serious market/message mismatch.
> Our care team has skewed the landing page to be a bit more of "show the benefit" rather than the functionality
That is what the snake oil industry does. Or enterprise sales. Even cults. ("Look at what we say these people say about us!" "We have a solution to your problem! [restated several times in different ways]!"
I am baffled by the term "care team" in this context.
I find that being concrete and credible, instead of asking people who don't know you for trust and unrewarded interest out of the gate, is a much better way to communicate something that is real.
If you do have a way to help ADHD people, I wish you luck communicating that. As an ADHD person myself, I have system creation/adoption fatigue. You seem to be aware of this. So be very direct about exactly what you do that helps, so someone that has tried many things, i.e. a sophisticated customer by necessity, can judge anything you say. (As they say in science, non-testable claims are not worth much. When marketing solutions to serious problems, this relates to the first thing you show people.)
Fair, feedback heard from multiple people here on: being more direct, concrete, credible. I'll take this back for the next iteration of the landing page!
Reputation for reliability, stability, or any other desired dimension.
Constant visibility in the news (good, neutral, sometimes even bad!)
A consistent attractive story or narrative around the brand.
A consistent selective story or narrative around the brand. People prefer products designed for "them".
On the dark side: intimidation. Ruthless competition, acquisitions, law suits, reputation for dominance, famously deep pockets.
To keep someone is easier. Tiny things hold onto people: An underlying model that delivers results with less irritation/glitches/hoops. Low to no-configuration installs and operation. Windows that open, and other actions that happen, instantly. Simple attention to good design can create fierce loyalty, for those for whom design or friction downgrades feel like torture.
AI-based product that slips past the defenses of people who think they hate AI, get turned off by branding like Copilot + PC, etc. A lot of people are really hoping it all dries up and blows away the way NFTs did.
Or maybe the honest to God non-dull tool that has nothing to do with AI. Like a Photoshop clone that does everything in linear light, makes gorgeous images, and doesn't crash when you open the font chooser.
In the context of optimizing parameters of a model, the Gradient consists of all the derivatives of the output being optimized (i.e. the total error measure) with respect to each of the models parameters.
This creates a simplified version of the model, linearized around its current parameter values, making it easy to see which direction to take a small step to move the ultimate output in the direction that is desired.
And easy to see which parameters adjust the desired output more vs. less.
The Hessian consists of all 2nd order derivatives, i.e. not just slope, but the curvature of the model, around the current parameter values.
Calculating all the first and 2nd degree derivatives takes more calculations and memory, but allows for more information as to which direction to take a learning step. As not only do we know how the output will respond linearly to a small parameter change, but whether larger changes will produce higher or lower than linear responses.
This can allow for the calculation of much larger changes to parameters, with high output improvements, speeding up training considerably, per training step.
But the trade off is each learning step requires more derivative calculations and memory. So a conducive model architecture, and clever tricks, are often needed to make the Hessian worth using, on larger models.
Another derivative type is the Jacobian, which is the derivate of every individual output (i.e. all those numbers we normally think of as the outputs, not just the final error measure), with respect to every parameter.
Jacobians can become enormous matrices. For billions of parameters, on billions of examples, with 100's of output elements, we would get a billions x 100's of billions derivative matrix. So the Jacobians calculation can take enormous amounts of extra computation and memory. But there are still occasions (much fewer) when using it can radically speed up training.
[EDIT] NxQxM 1st derivative matrix, N = #parameters, Q = #samples, M = #output elements
At this point, we have enough computer power and memory available, that all small enough problems should be trained with Jacobians in my view. Levenberg-Marquardt is an optimization algorithm that uses Jacobians. It can be orders of magnitude faster than gradient descent.
If I understand it in a nutshell. If Gradient is the angle Hessian is the curvature.
and Jacobians let you know how much weights contributed to the blue component of something identified as a big blue cat.
I think.
Jacobians look like they could be used to train concept splitters. For instance if an LLM has a grab bag of possible conversation paths, the final embedding would have information for each path, but once the selection is made it could filter the embedding to that path, which would be beneficial for chain of thought using the filtered embedding instead of the predicted token. I always wondered how much the thinking in embedding space carried around remnants of conversation paths not taken.
You explain well so what I never understood is how the Jacobians aren't the first derivatives themselves?
Also if you have happen to have any suggestions for linear algebra for someone who uses it without really understanding it (I can write a measurement function for an EKF from scratch OK, but I don't really understand why the maths does what it does) I would really appreciate it.
The Jacobian is first derivatives, but for a function mapping N to M dimensions. It's the first derivative of every output wrt every input, so it will be an N x M matrix.
The gradient is a special case of the Jacobian for functions mapping N to 1 dimension, such as loss functions. The gradient is an N x 1 vector.
[EDIT] Updated original comment to include matrix dimensions.
If you want a serious text that goes through the relevant linear algebra and optimization mathematics in depth up front, Neural Network Design, 2nd edition is a good one. [Disclaimer, co-author]. We took great pains to walk through every conceptual and mathematical topic before we apply those concepts to machine learning. We use MATLAB a lot, which may or may not be helpful.
Another potential option is "Linear Algebra and Optimization for Machine Learning", which looks good and also starts out with linear algebra before machine learning. I haven't read it, but the first 2020 edition gets good reviews, and a second 2026 edition just came out, apparently with a fair amount of positive revision. Given the speed of change, that's nice to see.
If your Windows / macOS / iOS / Android / Linux isn't doing Z right, just move to (whatever happens to be my favorite) Android / Linux / Windows / macOS.
Such ultra-reductive advice, in any arena, personal, technical, professional, is as as low quality as mouth babble gets.
Decisions that involve many trade offs and hurdles, both obvious, and specific to individual's circumstances, are not helped by annoyingly illiterate advice.
I wouldn't mind an HN rule aimed at curbing this kind of comment, which unfortunately, comes up regularly.
I thought the same thing, ... where is the "and Greenland!"?
The amount of hard, soft and economic power that are being burned for the bedtime stories of one person is unreal. As are all the cooperators and lobby harnessing conspirators whose actual dreams are getting implemented.
It isn't the fall of the USSR, but it is still a dramatic ceiling bounce.
Anyone using the Vision Pro as a virtual Mac screen, with environment bonuses, is someone who would like the option to drag Mac windows for powerful apps off their "Mac screen" and be their own windows.
And those Mac apps would be even better if in "Mac" windows on Vision Pro, I could have 3D models, an interface taking advantage of 3D for plotting, etc.
Instead, Apple believes the Vision Pro is an iPad Minus, instead of being the most powerful working machine they have. More Pro, than Mac Pro.
Sure, produce a "Vision" or "Vision Air" that is basically an iPad with 3D for the low end. But please let the "Pro" in Vision Pro fulfill its potential as the most powerful environment of them all. Stop holding back its potential. The iPhone and iPad need things simplified for their simpler interfaces. But Vision Pro's interface, with keyboard and trackpad, far exceeds a Mac's interface potential. Let' go, Apple!
Importing Mac-level capabilities, even up to the computing power of MacStudio Ultras, but only as a 2D screen in a 3D environment, just screams for so much more. The new top-capability Apple interface/device.
And a new level of computing/interface capability, will pull in the creatives/developers who will go on to add even more use cases. Instead of badly porting iPad apps.
Completely agree with Ben Thompson. Apple doesn't understand the potential of their own product, even as the hardware exists today, or last year.
Well you don't have to use the touchscreen, which was always rather awkward for many tasks, particularly anything productivity oriented. Even on my laptop i basically never use windows, but I still find the ipad to be unusable for anything but browsing and reading books and scrolling.
I'm sure they can do more. I'm just not sure "more" has anything to do with dragging windows around.
But maybe it's just me; I think it's easier to command-tab (or opt-tab) around than use my eyes. Unless I need to drag something between windows, which only happens about once a month or so.
> I'm sure they can do more. I'm just not sure "more" has anything to do with dragging windows around.
You get one Mac "window", with all your other windows jammed into it, as is necessary for physical screens, but entirely unnecessary for real 3D space.
Those "windows" could be frames containing three dimensional graphs, design structures, the list goes on and on.
The combo of eye selection, gestures + keyboard and trackpad are already wonderful. They need a little smoothing out, but work great already.
The contrast in power between Mac-in-Vision vs. the iPad-like-apps in Vision is dismal. The former is power but artificially flat and constrained, the latter is toy but with huge interface potential. The obvious software imposed (not hardware) gap in the middle, when using both, is enormous.
The device is made for so much more. As in, real computing in the "spacial computing". Otherwise, it is just a toy or an expensive Mac screen replacement. A waste of what should be a huge new category win for Apple.
Man, I just want to code from a more relaxing neck angle. I don't think all this is going to be very useful for that. I'm sure these ideas will be useful for something, but.... people have been talking about minority report for decades and trying to materialize it always ends up wasting attention processing too much at once.
For whatever reason, Apple delivered the hardware. But kneecapped the software.
They seem to see everything since the iPhone as a kiosk first, computer second. Which makes some degree of sense, Mac with keyboard/trackpad assumed vs. iOS/touch (with optional keyboard/trackpad) and covering many smaller sized screens.
But the Vision has much wider interface and interaction potential than Mac, not less. If Apple would let the Pro version of Vision, be a Spacial Mac. The new power high end.
(With a Vision Air, for the iOS-like, consumer, lower end market.)
Well, how else are they going to keep up their services revenue? Because at the end of the day, that's why they created the walled garden in the first place.
Orators learned the "palace of memory" trick for remembering long speeches. In that same vein, then, it does seem less demanding to simply be able to see where you put things.
Whether that's done by walking around, or just by glancing around on a 3D overlay (as suggested above for the Vision Pro), I like neither to have to search through stacks or folders of icons, nor to use Spotlight search fields. But perhaps the different types of cognitive loads result in what some people call different personal organizational styles or preferences. The "Clutterbug"[0] quadrant taxonomy comes to mind.
There really is an amazing untapped space of ideas on how to better navigate information.
Even in 2D interfaces, simple things: folders that looked fatter log-relative to how much they contain would add useful context and associative cues, and positive/subjective feelings of "real" recognizable locations vs. just a recursive "interface", when tap-tapping through folders.
An idea I implemented:
I hide a ".home" (zero byte) file in macOS folders I view as being at the top of a folder hierarchy. Then created a button in the finder toolbar that looks like a house. I can drill down a few folder layers, then pop right back to the hierarchy top by clicking the house button.
Just a simple thing. Ordinary users would understand the value of designating "home" folders. And once you have it you can't live without it.
For 3D:
I think traversable "Spaces" on screens were a great interface idea, done half way. and ripe for 3D extension. A space should be something that can be named, opened, closed, opened on another synced device, opened two years later. Duplicated or branched. I.e. a living persistent active project state of an open work state. With sub-spaces, for sub-projects, that can quickly be zoomed in and out of.
The latter would magnify the benefits of working on many different projects in a 3D environment, where having many things open and visible is really helpful, but laborious to continually reconfigure.
How nice to go into a rabbit hole on something important but not urgent, and be able to come back a year later to the same information still visibly organized where you left it. No context lost.
If there is an obvious "Minority Report" type power-user interface to be had, it would be that. Quickly navigating between presistent project/activity interface layouts with gestures. High value, high friction removal, with very low-bandwidth user direction needed.
Agree, it would be awesome to move all your mac windows all over the place instead of being constrained in a virtual display. That would cut the amount of displays you need o have in order to be able to see more without switching between stacked windows, I will be tempted to buy one, if that were possible.
I have 3 displays and I am tempted to replace them with a single one, but bigger (with something like the Samsung Odyssey Ark).
The line is if the symbols/works are used in a context so they clearly intentionally, or by unnecessary/unreasonable lack of care, create confusion. Someone who looks and sounds like McConaughey just being themselves isn't a violation.
Look at existing trademarks. They are riddled with high similarity filings, but they co-exist as long as they are not used to create confusion.
The bar for any enforcement would be very high for humans, simply looking and behaving like themselves.
But if someone very much like McConaughey was used in a commercial portraying a fictional "famous actor", that wouldn't go over. Unless ... it was clearly a parody. Or in fact, they are also an actor, and small signals indicate which actor, avoiding reasonable confusion problems. Or any other reasonable mitigations are taken.
McConaughey couldn't even sue a movie about him, with reenactments of real incidents in his life, using an actor naturally/made-up to look nearly indistinguishable, as long as it was clear the actor was not McConaughey. (Using computers to create an exact likeness might be challengeable, depending on the specifics - as they would essentially be lifting his face directly from him. Which gets into the realm of unreasonable, because it wouldn't be a reasonable requirement of any bio to go that far.)
> Every Country Should Set 16 as the Minimum Age for [Manipulation] Media Accounts
FTFY.
That is the real problem, no? The combination of surveillance, analysis of the surveilled data, very active feed manipulation based on that surveillance, and indirect business models that both finance and direct the specific manipulation.
Kids should be social. They should connect.
I think we do a grave disservice to our ability to reason about online safety by letting "social" be applied to what is largely interaction with adversarial/amoral value extracting algorithms, model-in-the-middle intermediating human connections, as if the result was any kind of natural social behavior.
Whatever artifact makes sense. Maybe a spreadsheet list. Maybe a one page, or half page, serious coherent characterization from each person.
How people independently see priorities before collective discussion, being as important as how they rate them, I would go with one page thoughtful summaries.
Having experienced partially-aligned organizations, I would want key contributors to a discussion like this to come from every major part of an organization. Emphasis on very different roles, especially important leaf roles, not just top leaders. Otherwise you get high bubble-at-the-top alignment. Top-only strong alignment is like concrete, perniciously inflexible. Which is what I experienced. End-to-end insight and discovery of alignment is needed to get adaptable, in-touch, actually effective alignment.
And as is hammered so well in "Creativity, Inc" by Ed Catmull, alignment isn't just about coordinating efforts. It is about continuously identifying and removing friction, of any kind, anywhere, for anyone, as individuals make their very different contributions in service to the common direction.
reply