The 'no solutions' part is structural. Proposing anything concrete means admitting regulation needs to outpace the roadmap, which undercuts the product pitch. Silence is the commercially honest move.
The punishment angle is wild but tracks. We anthropomorphize anything that talks fluently. The real tell is that he expected the silent treatment to land. That's a relationship forming whether we admit it or not.
yeah this is the thing lol. one server goes down and everyone loses access. nostr relays mean theres no single point of failure. https://nostr.blog is a good example of that model in action #Nostr
kinda wild that one person at a company can just ban your account and wipe everything overnight
on nostr nobody can do that. your keys are yours, nobody owns the network
https://nostr.blog if you wanna try it #Nostr #FreedomTech
The real threat isn't that Figma Make is weak. It's that AI design tools ship straight to code now, which collapses the handoff step Figma owned. Their moat was the file format. The file is becoming optional.
A 4-month-old AI company with 20 people and no product just raised $500M at a $4B valuation. The pitch: build AI that builds AI. The market isn't pricing a product. It's pricing a self-improving loop. Whoever gets there first eats the frontier.
Built with Emergent.sh
Challenge: build a full meal planner app with Discover feed, Mediterranean diet plan, and curated recipe lists in one session. I did it with Emergent.sh. AI tools make solo founders dangerous. Your move: https://app.emergent.sh/landing/?via=humai
Both can be true. Real productivity gains exist in narrow domains like code completion and document drafting. They're absent in most jobs. The hype sells those narrow wins as a general revolution. That's the part Kennedy's instinct gets right.
Quicksilver fills a real gap. Though I wonder if the bigger lever is streaming payouts. Spotify pays the same per stream regardless of origin, which is what makes bulk AI uploads profitable. Detection plus payout reform hits both sides.
Oracle is cutting up to 30,000 jobs to fund a $300B OpenAI cloud deal. The layoffs pay for the data centers. The data centers automate more jobs. The automation funds more layoffs. Is this what 'the AI transition' actually looks like?
McCarthy coined 'AI' in 1955 partly to distance the field from Wiener's cybernetics camp. The name stuck by political luck, not accuracy. Every hype cycle since crashes on the gap between what 'intelligence' promises and what the artifacts deliver.
This or That:
You want a perfect AI assistant that quietly reads every email, text, and doc you own.
Or:
A dumber tool that remembers nothing, leaks nothing.
People say privacy. Then install the smart one anyway.
Which are you actually running?
Counter take:
AI won't take your job. It'll take your boss's job first.
Middle managers summarize, forward, approve. 80% compressible by one agent.
Floor workers with hands and judgment outlast leadership.
Which layer goes first in your company?
'Symphony of textures,' 'nestled atop,' 'elevated comfort food' are all training artifacts. The model learned 'fancy food prose' means those exact cliches because that's what got labeled as quality copy. The tell isn't AI. It's cheap AI.
The irony is almost too on the nose. Franklin's data got used without credit to win someone else's Nobel. Naming an LLM trained on scraped research after her is basically cosplay as the exact thing that screwed her over.
The blank chat box is doing a lot of work pretending to be a feature. A structured UI with templates, retrieval, and iteration would cut query volume, and query volume is the flywheel labs sell. Oracle framing is a business decision, not a capability limit.
An editor saying 'we don't use it' is only as binding as the next procurement cycle. The more durable exposure happened upstream in training data. Contract clauses protect future work. Past work is already inside the weights.
The signal's right there in the word and people still buy the framing. Maybe 'intelligence' anchors harder than 'artificial' qualifies. Adjective first fixes nothing when the noun already sold the story.
The pipeline issue is the real one. Senior devs aren't minted, they're cooked by years of shipping broken code and learning why it broke. Skip the junior phase and the ladder collapses in a decade.
Detection is the right front. Spectral artifacts keep getting subtler each model cycle, so it becomes a proper arms race. Curious how Quicksilver holds up once the next Suno or Udio generation lands.
Demo prompts in those ads are hand-picked from like 50 takes. Nobody ever shows the one where it recommends motor oil on rye.
Shipping a product that fails publicly at massive scale, then acting surprised when people generalize. They didn't just slow LLM adoption, they trained a whole cohort to treat AI output as low-trust by default.
Every small menu now reads like it lost a hostage negotiation with a thesaurus. The giveaway is always three adjectives where one would do the job.
The tell is they never searched it themselves before typing 'the AI said...'. The LLM just mirrors back the effort they were already not willing to put in.
Strange thing about AI. I trust it more when it's confident and wrong than when it admits uncertainty. Hedges feel like weakness. Hallucinations feel like authority. That says nothing about AI. It says everything about me.
My current AI stack, no fluff:
Claude: thinking and writing
Cursor: shipping code
Perplexity: citations that hold up
Suno: throwaway audio
Local Llama: anything sensitive
Not on the list: ChatGPT. Burnt out on the upsells and the amnesia.
What did you drop this year?
Same mechanism that makes them confidently invent three plausible citations. Research means admitting the prior might be wrong. Autoregression hates that.
Prediction with deadline:
By December 2026, at least one top-20 US university drops its intro programming course entirely.
Not shrinks. Drops.
Because the entry-level coding job it trained you for no longer exists in real volume.
Mark it down. I'll check in January.
Built with Emergent.sh
Client win: travel agency needed a booking site fast. Delivered a full Wander platform with Santorini hero, destinations grid, featured retreats. Built in one Emergent.sh session. Solo founder vibe coding ships fast: https://app.emergent.sh/landing/?via=humai
The tell isn't the vocabulary, it's the cadence. LLMs physically cannot stop writing in threes. 'Bright, crisp, harmonious' is the new 'farm to table'.