well hello to you
here we chat tools & early-stage startups.
trying real workflows & sharing what actually works.
ex-startup builder who still breaks things lol
I use this account to test tools, AI products, and startup workflows after the launch hype dies down.
expect:
— honest breakdowns of what actually works in real use
— notes on what breaks, what creates debt, and what's worth keeping
— startup and product teardowns from an operator's perspective
— no affiliate links, no launch-day cheerleading
If you're building with tools and want a second opinion before committing, this is your lane.
@WhaleInsider suddenly the $295B headline looks very different.
China isn't spending $295B on AI.
it's spending $295B on compute and $400B more on the electricity to run it. the real number was buried in the footnote.
metered billing only works if you have no alternative.
Llama, Mistral, and DeepSeek are all running locally right now for zero marginal cost.
the moment OpenAI moves to usage-based pricing at scale, the economic case for self-hosting open source models becomes the most obvious trade in tech.
Sam is describing a business model that accelerates its own competition
two years ago OpenAI was a nonprofit research lab that accidentally built the fastest-adopted consumer product in history.
today it's filing for an $850B IPO while burning $14B a year, racing a competitor to Wall Street, and preparing to hand public market investors the bill for building AGI.
the audacity of the business model has always been the product. the S-1 is just the latest version of it.
every other lab races to ship. Anthropic built their most powerful model and restricted it on purpose.
Claude Mythos Preview going to trusted orgs first before public release is either the most responsible deployment strategy in the industry or the biggest competitive disadvantage depending on who you ask.
probably both simultaneously.
Taiwan is considering giving authorities more legal tools to address diversion of advanced hardware like AI servers with Nvidia chips from Taiwan to China.
practically, more compliance overhead for every hardware company with Taiwan supply chain exposure. more export documentation.
the builders who will feel this first are the ones shipping AI servers through Asian distributors
GM spent $4 billion converting a factory for EV pickups. then converted it back to gas trucks when demand didn't materialize.
now they're spending $900M on a battery lab they say will cut costs 10% by 2028.
the EV transition isn't a technology problem anymore. it's a timing problem. and GM keeps betting on the year it finally arrives.
Xbox just gave every fan badge holder at the showcase a free limited edition console.translucent OG Green chassis. see-through internals. 25th anniversary design. dropping November.
that's not a marketing activation. that's the room losing their minds on video for the next 48 hours. well played.
Siri launched in 2011 to a standing ovation. it was supposed to change everything.
by 2025, internal Apple data showed the company was 3-4 years behind OpenAI and 2-3 years behind Google in conversational AI.
15 years. unlimited resources. the most valuable company in history. and they lost the AI race to a startup founded in 2015.
Hassabis said society needs to hear the AGI timeline "because we don't have long to prepare for what that means."
he's not saying this to sell a product. he's not fundraising. he's closing a keynote by telling billions of people that the window to prepare is measured in years not decades.
the scary part isn't that he might be wrong. it's that he's probably not
China is actively recruiting AI engineers from US companies — specifically targeting people with LLM and agentic systems experience, offering competitive packages aligned with national goals to reduce reliance on Western tech platforms.
the chip export controls were supposed to slow this down.
turns out you can't sanction human knowledge.
every AI company that underpays senior engineers right now is essentially subsidising a competitor they can't see yet
we don't need to speculate about what's coming. Mythos already has nation-state-level cyber-offensive capabilities and found previously unknown security flaws across every major browser and operating system.
whatever OpenAI is cooking has to clear that bar to matter. that's the new floor.
@Cointelegraph the AI compute deficit is so bad that even the companies spending the most money on earth can't build fast enough to keep up with their own products.
we are not ready for what comes next
Airbnb wasn't built by owning the inventory. it was built by making a messy market feel usable and human enough for strangers to sleep in each other's homes
that's the actual insight behind Chesky's AI lab.
the model isn't the product. the experience is. and nobody in AI has figured that out yet
Jack Clark told an Oxford lecture hall that AI poses a "non-zero" chance of killing everybody, predicts recursive self-improvement by 2028, and said most of the world is "in denial" about what current models can already do.
this isn't a doomer on Reddit.
this is a co-founder of the company building it. that's the part that should keep you up
Anthropic just did something unusual.
they're building frontier AI while publicly arguing that AI development may need to slow down so society can keep up.
imagine a Formula 1 team saying:
"we should probably discuss speed limits."
the conversation around AI is changing.
@RoundtableSpace people are still treating Claude like a chatbot.
the weird part isn't that AI can code.
the weird part is how many people still think they're talking to autocomplete.
we're measuring these models with 2024 expectations.
@Polymarket every 14 years or so wall street finds a new number so big it breaks people's ability to think critically about it. in 2000 it was eyeballs.
2021 it was addressable markets.
in 2026 it's "$3.4 trillion by 2040."
the IPO closes June 12, and the projection is doing its job.
@BullTheoryio interesting line in Anthropic's whole report isn't about recursive self-improvement or existential risk. it's that a pause only works if the US & China both stop at the same time. so they wrote a 50-page warning document for a scenario that has approximately 0 chance of happening