worthless ux + .1% wisdom

Z / Runner 🏃🏻‍♂️

🏃🏻‍♂️ digging deeper (death of software 2.0): "human-oriented consumption software will likely become obsolete". "Faster workflows, better UIs, and smoother integrations will all become worthless". "intelligence is dynamically assembled at runtime using retrieval (RAG), external tools, and active context management".

• cross-system orchestration w/out custom APIs.

• tracking drift (ideas, markets, health, narratives) over time.

• accumulating judgment (retiring heuristics, updating beliefs).

• 🏃🏻‍♂️ perspective matters; hammers only see nails; screwdrivers only see screws… too early to pronounce the death of UX/UI… just like most folks think google is dead on ai… the useful bits about the article it is a counterfactual to not build a dumb/static product.

• 🏃🏻‍♂️ doubt it, finance is already oversaturated w/ quant heavy + reward hungry seekers; unlikely a general agent will outperform hyper-targeted ai (see below quote from investment banking "can through a bunch of PE's and call out any of their portcos that had certain characteristics-the ultimate goal was to create a buyers list for a sell side mandate. In the olden days, I had to go on each of their websites and look at each portco to find ones that matched what we were looking for, then dive into Pitchbook and doublecheck, and so on. Scanning 50 PE firms with 20 investments each took hours to come up with a list of 10-15 to reach out to. This took ChatGPT 2 minutes to pull together websites for all the portcos, and then it took the intern another 20 to go on the websites and confirm the information GPT said was accurate. The prompt was basically "review the portfolio companies of this list and return any that would be interested in acquiring a company of X capabilities and Y size."

• 🏃🏻‍♂️ skeptical, ai making $$$ is real, a real way to get everyday folks excited.

• 🏃🏻‍♂️ too broad; taking the bitter lesson too literally, if you're everything for anything, then you may end up doing only something.

• 🏃🏻‍♂️ false debate, only <5 viable LLMs in 2026, all gunning for "AGI", everyone else, including claude is specialized agent. Hence, why the debate about general agent vs. specialized ai skew and warps b/c the definition and targeted application use cases are so important. For example, i never considered manus to be a general agent per my own & somewhat limited use cases… but i think you considered manus to be one.

• 🏃🏻‍♂️ "human-oriented consumption software will likely become obsolete".... Two possible paths (arduous but rewarding)... VLA/world model = physical ai (ie. robotics) or extension towards metahuman like abilities. In either case, i doubt browser and mobile will continue to play the crucial role. We should be mindful of falling into the trapping of our "aging" mindset. In the context of this topic, we have to unground our old ways of thinking about human in the compute loop.

Stephen / Basketball 🏀

• 🏀 hence, the death (on arrival) of your beautiful (journaling) ux?

🏀 top 10 categories of professionals / prosumers with high spending on ai tools?

• per claude + clio: software systems across specialized domains (10%), professional and technical content (8%), business growth and revenue generation strategies (8%), academic research and educational material development (7%), information about people, places, or organizations (5%).. the rest (60%) - such a long tail!

• "top 10 youtube channels for financial professionals": AswathDamodaranonValuation PBoyle ThePlainBagel BenFelixCSI MoneyMacro TheSwedishInvestor KenjiExplains rareliquid MichaelKitces markets.

• 🏀 my exact opposite takeaway: per peak's podcast, a general agent for specialized tasks for non-specialized prosumers (okbet + alpha arenas / $1b fund / $1m profit).

• 🏀 those who cannot do, _.

• ditto for _ professionals: finance, investment, business, consultant, revenue, growth, analyst, strategist, sales, marketing, legal, fintech, e-commerce; education, academic; people / place / organization / market research, recruiter; technical, engineer, developer, security, creative; basketball, ballet.

• 🏀 not broad enough for some. but easy dichotomy to settle: in 2026 there will be much more $100m general agents than specialized ai.

• 🏀 per "peak".

• 🏀 that said, generality acts as the floor; so unlock with local optimizations: mixture of experts in 2023, chain of thought in 2024, context kv caching in 2025, formal tool use in 2026.

🏀 #bitter-agent 1000x compute (search + learn) vs top .1% wisdom (design + code) – on general (video) ai / crawler. z, if your second choice (but 100% ai native) is vision-language (action) model, would you consider browser / mobile first as your bot (2d) playground / world?

• 🏀 definitely just a stepping stone to the matrix.