OpenAI offering the stake proactively makes 66% feel conservative.
the question was never whether the US government wanted in - it was whether OpenAI would make it easy enough to say yes.
they did.
@kimmonismus every frontier lab eventually decides the chip layer is too strategic to rent forever.
Anthropic just started that clock.
Samsung needed a headline customer for its 2nm push.
both problems solved in one announcement.
going from "coding model" to "generally intelligent model" is a much bigger strategic shift than people realize.
that's actually a platform play, not a feature play.
New cursor model being teased at compile
- same size as Claude opus and gpt 5.5
- trained from scratch, no more kimi base
- 10-20x more compute vs composer
- generally intelligent, not just coding
Releases in the “Next couple of weeks”
Bezos saying AI creates jobs while simultaneously building a system to automate end-to-end engineering is the same tension Anthropic publishing RSI concerns while shipping Opus 4.8.
everyone believes it and nobody is stopping.
BEZOS SAYS AI WILL BOOST JOBS, LAUNCHES PROMETHEUS
Jeff Bezos said AI will create a labor shortage rather than destroy jobs, arguing it will boost productivity and expand employment opportunities. He is co-leading new AI firm Prometheus, which aims to build an “artificial general engineer” to speed up product design and manufacturing. The $41 billion startup has raised $12 billion and seeks to automate end-to-end engineering while increasing output from smaller teams.
@kimmonismus Anthropic approaching profitability while OpenAI considers price cuts to compete is the business model story that makes every benchmark comparison irrelevant.
the sustainable lab sets the terms.
the burning lab chases them.
the company building recursive self-improvement just published a document saying society isn't ready for recursive self-improvement.
and then kept building.
the gap between what they're writing and what they're shipping is the most important story in technology right now.
OpenAI just wrote: "We also see early signs of recursive self-improvement (RSI) in today’s systems: where AI development is itself accelerated by AI.
We expect this to increase competitive pressures among developers and nations, and create governance challenges that existing institutions are not equipped to address.
As RSI emerges, societies will need ways to shape the trajectory of AI development and ensure that it serves human interests."
The vibe has changed, something is happening.
@JulieLovesTech NVIDIA just dropped a 550B open weights model optimized for the exact use case every serious AI builder is actually deploying.
the benchmark era is over.
the "does it work when my agent runs for three hours" era just started.
this is the part people miss about mainstream crypto adoption: users rarely adopt “crypto products.”
they adopt better banking experiences.
if stablecoins become invisible infrastructure inside existing apps, adoption accelerates much faster.
the first nationally chartered bank to issue a stablecoin inside its own app just made every other bank's digital payments roadmap look three years behind.
15 million members didn't need to download a new app or open a crypto account.
it was just there when they opened SoFi.
that's how mainstream adoption actually happens - not with a product launch, with a feature update.
@JulieLovesTech the irony of AI is that stronger models often make human expertise more important, not less.
because the better the outputs look, the more dangerous subtle mistakes become.
meta building a Reddit clone while Reddit trades at $150+.
the timing is interesting given Reddit's stock run.
Zuckerberg has never seen a product category he didn't want to own.
Meta Platforms $META has quietly released a new standalone app for Facebook Groups, called “Forum.”
The company seems to be positioning Forum as a platform that functions similarly to Reddit - Tech Crunch
@perplexity_ai security tooling that's open source and auditable is the only kind worth trusting on a dev machine. Bumblebee connected to real-time threat intelligence is exactly the kind of infrastructure the ecosystem needed. good ship.
74,000 weekly tasks at PayPal alone.
Perplexity isn't winning the consumer AI war.
it's quietly winning the enterprise research war.
and nobody's paying attention because it doesn't have a viral moment every week.
that's exactly how category leaders are built.
PayPal runs 74,000 weekly tasks in Perplexity Enterprise.
Teams use it for model validation, channel performance, market trend research, competitive intelligence, and product analysis.
Read the customer story: https://t.co/2bzOjxvYee
@StockSavvyShay you're building an AI layer on top of the largest consumer financial network outside of banks.
watch what they do with merchant insights and fraud detection first.
Anthropic has 40% of its top customers in finance. now Perplexity is launching a dedicated finance product. Bloomberg has been building AI features quietly. the entire financial data and analysis industry is getting rebuilt simultaneously.
Today we're launching Perplexity Computer for Professional Finance.
Finance teams can bring licensed data from providers like Morningstar, PitchBook, Daloopa, and Carbon Arc into Computer.
We’ve also added 35 dedicated finance workflows for the work analysts repeat every week.
Google leaked an AI that watches your screen, remembers everything, and browses for you.
then deleted it.
things that aren't real don't need to be deleted.
Google just leaked COSMO...
> Local Gemini Nano
> Screen access
> Voice match
> Recall
> Browser agent
> Deep research
Then it vanished.
Android is becoming an AI agent👀
@dmitry140 traditional SaaS scaling: revenue doubles, headcount follows closely behind
Perplexity 2025: revenue 5x, headcount up 34%
same growth goal.
completely different human capital equation.