Rephrasing slightly: knowing when the thing is actually WORKING (to solve the problems / meet the opportunities the product is intended to be addressing) becomes the hard problem.
Ideally, writing the product requirements becomes an explorative learning process rather than a political / persuasive process: the PMs learn what features really solve user problems rather than argue about which might do so.
And if there's no good-enough open solution today, how does the open web community organize itself to build it? If the answer is "vibe code it with Claude" is that cure worse than the disease?
To start: What EXACTLY would you propose as an alternative to Instagram for a band wanting to publicize its tour? How would the kind of folks who can easily post to Instagram build it? How would you suggest they get the same kind of traffic to it as Instagram's algorithms do?
That is a profound question. I'd love to see the W3C AB broker a discussion of why the web user experience is not competitive with walled gardens like Instagram, and what better tech and narratives might change that.
Not a WHOLE lot more than back in the day when I wondered how much was written by under-credited subordinates, assistants, interns, students … or just cut and pasted from Google. The important questions are about utility, reproducibility, quality, etc not attribution
For better and for worse I suppose, hand crafted code will probably become like small-scale produce, handcrafted furniture, etc.: Appreciated by the knowledgeable, too expensive for the mainstream market, sustained by a subculture respecting beauty and craftsmanship over cost efficiency. 😔
It's a good analogy in another way: the "Luddite" rebellion against automation in the early 19th century was targeted at "stocking frames" that automated sock knitting. So hand sock knitting is very symbolic!
As someone who hasn't paid a dime for car fuel (or home electricity) since getting an electric car + home solar a year ago, I keep thinking it's only a matter of time before US consumers get the word: There's an upfront cost, but it's paid off fairly quickly in operating savings and peace of mind
I wonder what a poll of CEOs in 1985 would have said about the ROI of PCs, or the web in 1995, or smartphones in 2008? Those who hoped for magic didn't get ROI, those who figured out how to leverage their strengths / mitigate weaknesses to solve real business problems did.
Yup, the Google one. The noisy — both visual and content— MSN home page is what made me give up on Edge after retiring from MS
Well, "triage" means classifying ideas that are ready to be implemented, promising ones that need work, and those that aren't worth / ready for a lot of work. It's hard, and mistakes happen. But if you let things just muddle along until the choice is obvious, you get the current situation 🤷♂️
In a single org, there might be Product Managers that triage out not-quite-good-enough ideas, but the web platform doesn't have a Project Management team.
I've heard arguments that the TAG/ should evolve into one, but that idea isn't quite good enough to get traction either
"Good enough" ideas have a high threshold because "yes" means "we have to support it forever". For example client-side XSLT seemed like a great idea 25 years ago, but now it doesn't have enough users attract maintenance resources, but enough to make it difficult to remove. No-win scenario.
"Still, every year a huge number of things remain in the pool that we can’t afford to take up. The pool keeps growing."
That seems be the root of the problem. Nothing kills off "not quite good enough" ideas.
I probably had undiagnosed ADD a kid, and never was good at math. But tech such as slide rules (I'm OLD) and calculators liberated from the tedium to learn how to use math/stat/programming to analyze problems. I now use AI chatbots to liberate me from the tedious stuff and focus on the interesting
Agree! But chatbots aren't going away. Design home work to route around ways to "cheat"? Assume students will "cheat" and help them learn how to use new techniques to thing about problems? Somehow isolate them from the bots when assessing progress?? But help them learn how to learn.
II understand, but I lived through similar concerns about using calculators in schools, then computers, then the internet, long before AI. For me (a bit "neurodivergent") these were tools that motivated me to to learn the concepts without bogged down in the "friction". So, it's complicated 🤷♂️
If education can't adapt, the educators were poorly educated 🫠. Paper and pencil in-class exams may rise again....If AI chatbots help people LEARN, they will still get used. If not, 🤷♂️
As a browser PM, I went through the stages of grief about all the stuff that should be done in an ideal world, and came to Acceptance that "evolution is cleverer than you are" en.wikipedia.org/wiki/Orgel%2...
So the most used bits of SVG get maintained, server-side XSLT for static documents gets maintained, tooling gets created to make things like math usable by ordinary mortals using LaTeX or whatever.
The web platform evolved more than it was designed. There are multiple ways to do almost anything. So lack of attention to something marginal is evolution's way of focusing attention on what is important to getting work done / making money / having fun...
Thanks! My current obsession is wondering if it would be possible to pivot the LLM-AI industry away from "winner take all" into the less zero-sum mindset, like the early Web. I'm dreaming , but ... Could some org do what W3C did for the early web? W3C AB topic now that you and Dan are on it?
The most compelling being opportunity cost ... any engineers skilled enough to to implement XSLT could be used more lucratively on lots and lots of of other things
But ahh, the memories this brings ... my very first meeting as a Microsoft employee in 2004 was about whether to upgrade to XSLT 2.0 in the native (used by the browser) and .NET XML Libraries. IIRC we decided "YES" but it never got done for a million boring practical reasons.
So styling RSS is the main use case for XSLT in browsers these days? Seems like a large hammer to swat that fly 🫠
The practical challenge is to build human-machine systems that amplify human emotional/value judgment with automated pattern matching ... to play better 3 dimensional chess against the stupid people and machines that threaten the world in so many ways.
LLMs have demonstrated that a lot of tasks that we assumed required a high level of "general intelligence" were mostly pattern matching, e.g. coding. And recent events have shown that supposedly intelligent humans can make horrible decisions driven by emotionally satisfying self delusion
Sad but not surprising to year, 6 years after retirement means I don't have to dread Connect Season any more 😱
"growth mindset" was a constant message from Satya in his early days as Microsoft CEO, certainly helped create the resilience that drove its continued success as tech and business models changed. When did it morph into "you can be replaced by an AI mindset?"
