rami context

here’s the thing, AI news is easy to report. actually understanding it? this is where I break it down — without the hype. notes on: — what actually changed — what it means in real workflows — what people are over/underestimating for the girlies in AI (and the ones getting into it) who want signal, not noise.

@GenAIDL the winner may not be the one with the largest models, but the one that uses them best Americans built the best tools on earth and then argued about whether to use them

The US is winning the AI model race. China is winning a different race and that one might matter more. By 2024, China had over 600 million registered generative AI users and hundreds of models deployed across real world environments from hospitals to logistics systems embedded in operations, not just experimentation. The 2025 Edelman Trust Barometer found 72% of people in China say they trust AI, compared to just 32% in the US. Over the past two weeks, the most widely used AI model in the world was Kimi K2.6 an open-source Chinese model that most Westerners had never heard of. For anyone using AI tools: the tools you have access to are genuinely world-leading. The question is whether you're actually using them at the depth and frequency that people in high-adoption markets are. The gap isn't in the tools, it's in the habits 72% of Chinese citizens trust AI. In the US, it's 32%.

AI is being used to find hacks. AI is being used to stop hacks. AI can now reduce the time between finding a vulnerability and exploiting it dramatically, while making the attacks harder for traditional scanners to detect your 2FA isn't the problem anymore. the problem is an AI that reads code better than the dev who wrote it

Google stopped a zero-day hack that it says was developed with AI theverge.com/tech/928007/go…

@colepulse louder for the people on the back

The "AI is a bubble" takes always cite the same stat right now: 41% concentration, same as past bubble peaks. The stat is real. The conclusion is incomplete. What's similar to 2000: Concentration above 40% Capital flooding into infrastructure A small group of stocks driving the entire index What's different from 2000: Today's leaders are profitable. Most dot-com leaders weren't. Today's P/E ratios are 1/4 of 2000's most-extended levels. AI infrastructure is generating revenue NOW, not promised future revenue. What this means: the bubble framing is wrong, but the concentration framing is right. Those are different things. A market with concentrated leadership AND real underlying earnings can still draw down 30-40% if the earnings growth slows or one major player stumbles. That's not a bubble bursting. That's concentration math. The honest position: bullish on the technology, cautious on the concentration. Both can be true. Most takes pick one and dunk on the other.

@pubity we’re moving from “people making podcasts” to “systems generating podcasts” AI voices talking to AI voices, producing hours of episodic content with no studio, no schedule, no fatigue the cost of audio content is collapsing toward zero attention is now the scarce resource

Around 35% of new podcasts are completely AI generated according to the Podcast Index. Episodes feature literal hours of AI-generated voices talking to each other, and AI firms are generating thousands of episodes with almost no human involvement.

@Polymarket this is what convergence looks like in real time AI stops being a separate “tech category” and gets absorbed into aerospace, communications, and infrastructure platforms “SpaceXAI” isn’t a brand shift it’s a stack consolidation move

NEW: Elon Musk announces xAI will be dissolved as a standalone company & will instead be folded into “SpaceXAI”

the Pentagon isn’t “using AI tools” anymore it’s wiring AI into the process of how decisions get made Scale AI handling data + decision support across military workflows is less about software adoption and more about delegation at scale once decision support becomes automated, the system stops being human-first by default

The Pentagon awarded a $500 million contract to Meta Platforms Inc.-backed Scale AI to help sift through data and assist in decision-making, the latest step in the US military’s effort to rely more on artificial intelligence. bloomberg.com/news/articles/…

@infranotes if your system still requires humans to orchestrate steps, you’re not building the future stack — you’re simulating it.

@aicontext_ stop worrying about model smarts. the real edge is multi-agent orchestration ngl. if you are still building workflows for humans to run manually you are already behind the curve.

The next AI shift isn't smarter models. It's AI agents working in teams. The first wave: agents that browse the web or write code — alone. The next wave: agents that collaborate, delegate, and solve complex goals together. Google just launched its Gemini Enterprise Agent Platform. Five Eyes intelligence agencies have issued joint guidance on agentic AI in critical infrastructure. JPMorgan is deploying AI agents to scan $10 trillion in daily transactions. Agentic AI is no longer experimental. It's the new enterprise infrastructure. The real question: are you building workflows that humans run, or workflows that agents run with humans in the loop?

@tcloudlabx AI is starting to look less like a model and more like an operating structure.

@aicontext_ google just dropped the gemini enterprise agent platform. agents are moving from solo browsing to full team collaboration.

@learncurvesai and they’re moving at completely different speeds.

@aicontext_ five eyes issuing warnings while jpmorgan points agents at ten trillion dollars. what could possibly go wrong

@Forbes Anthropic didn’t win by being louder it won by getting embedded into actual workflows. Claude Code took developer mindshare. finance agents are starting to take analyst workflows next. distribution inside enterprise stacks matters more than Twitter hype.

Anthropic Is Now Targeting Finance After Revolutionizing Coding forbes.com/sites/the-prom… (Photo: Getty Images)

all these companies agreeing on an open networking protocol is not a technical story, it's a political story these five companies agreeing on anything means the alternative — every hyperscaler building proprietary networking stacks — was hurting all of them more than competing with each other was helping open standards don't happen from generosity. they happen from shared pain.

We’ve partnered with @AMD, @Broadcom, @Intel, @Microsoft, and @NVIDIA, to release Multipath Reliable Connection (MRC), a new open networking protocol that helps large AI training clusters run faster and more reliably, with less wasted GPU time. openai.com/index/mrc-supe…

the Pentagon just signed AI deals with OpenAI, Google, Microsoft, SpaceX and more Anthropic? blacklisted — because they refused to let the military use Claude for autonomous weapons that's a wild position to be in: turning down DoD contracts worth billions because of your values whether you agree or not, that's a level of conviction most companies will never have

@sksk_sinker the weird part isn’t yields being higher it’s that Treasuries aren’t stepping in when equities wobble that used to be the whole point… now it’s just not showing up consistently

The headline yields are real. The bigger story is what they're saying about the bond market's role. 10Y at 4.45%. 30Y at 5.02%. What's actually new: Treasuries are not behaving like a safe haven during this Middle East conflict. They're trading like a risk asset. Why: oil-driven inflation makes bonds the wrong hedge. When the shock comes through energy prices instead of demand collapse, fixed-income payments lose purchasing power before equities lose earnings. For portfolio implications, this is the part to sit with: The "60/40" assumption (stocks down, bonds up) is broken in this specific scenario. If oil stays elevated and yields stay around 4.5%, hedging equity exposure with Treasuries does not work. Gold, energy stocks, and commodities are the substitute hedge - and they have been outperforming Treasuries since the conflict started. Whether or not we hit 5% on the 10Y or 7% on mortgages depends on factors nobody can position around. What you can position around: knowing your hedge stopped working in March.

NEW WEEK: if OpenAI is building a phone, the interesting part isn’t “no apps.” it’s the shift to continuous agents that sit between you and every action on the device. that raises some uncomfortable engineering questions: • when context is wrong, who is responsible? • when the agent acts incorrectly, how do you unwind it? • when it stalls, what does the UX look like? how much compute does “always-on understanding” actually cost? we’re used to apps failing. we’re not used to agents failing on our behalf.

@BusinessInsider imagine being a KPMG consultant right now. 4% of the US advisory business just got cut. and for the people who stayed, AI usage is now literally tracked and baked into performance reviews. use AI more or get managed out. very normal employment relationship we’ve built here.

KPMG says it hopes the dashboard encourages more "frequent and sophisticated" AI use among its US advisory's 10,000 workers. bit.ly/4n4It8Y

@MikeHapner @unusual_whales also add fuel to the short list before penthouses

@unusual_whales Can we just have affordable food in the meantime?

Elon Musk has said: "AI/Robotics will mean everyone can have a penthouse if they want. The output of goods & services will be several orders of magnitude higher than today’s economy."

people keep reacting to the OpenAI phone idea like it’s “another smartphone.” but the argument is really about data access. the best AI assistant is probably the one with the deepest real-time understanding of your context, behavior, routines, and intent. today almost all of that lives inside ecosystems controlled by Apple and Google.

Anthropic is partnering with Blackstone, Hellman & Friedman and Goldman Sachs to form a new company that will bring artificial intelligence to a broad group of midsize companies bloomberg.com/news/articles/…

@garrytan the most interesting AI lab right now might not be a lab. it might be a repo.

In the future everyone will have a personal AI This is why I'm open sourcing the code and skills for mine semafor.com/article/04/24/…

@MorningBrew $4B to $900B in 36 months. no ads. no consumer hardware. no search engine. no social network. golden

Anthropic's valuation over time: • May 2023 $4.1 billion • Feb 2024 $18.4 billion • Mar 2025 $61.5 billion • Sep 2025 $183 billion • Jan 2026 $350 billion • May 2026 $900 billion(?) That escalated quickly