Tech & AI Twitter Unfiltered: Power Plays and Shifting Alliances

An internal memo from OpenAI just accused its partner, Microsoft, of acting as a "gatekeeper" that limits its reach to enterprise clients. This is the first public crack in the most important partnership in AI, and it signals that the fight for control over the future of this technology is now moving from the lab into the open market. This week, the conversations on tech Twitter are moving past the simple awe of new models. They're about power, money, and the fundamental nature of the progress we're all witnessing. The themes are consolidation, control, and a growing, serious question about what this all actually adds up to.

For instance, Google's DeepMind just hired a philosopher from Cambridge, Henry Shevlin, who will start in 2026 to study AI consciousness and ethics. This isn't a PR move. It follows a pattern set by Anthropic, signaling that the top labs are now treating the moral and philosophical risks of advanced AI as an engineering problem that requires a dedicated, senior expert. They are building their conscience in-house. At the same time, the UK's Competition and Markets Authority is now raising alarms about a "two-tier" AI economy. Their analysis makes a sharp distinction: consumer-facing AI is everywhere, like using a public road.

But access to the frontier models, the massive compute resources, and the deep integrations that actually create economic value... that's like owning the toll booth. That access is becoming more concentrated, not less. Transparency on what data these models are even trained on has fallen, with the industry's score dropping from fifty-eight out of one hundred in 2024 to just forty this year. The curtains are being drawn. And this leads to the most provocative conversation of the week, bubbling up from a thread on Hacker News. The question isn't whether AI is powerful. The question is whether AI is the start of a new wave of innovation...

or the final, frantic end of the last one. The argument, based on the work of economist Carlota Perez, is that AI isn't creating a new economic reality. It's just optimizing the digital one we've been living in for thirty years. The evidence is compelling. AI investment peaked at two hundred billion dollars in 2025, but economy-wide productivity gains have stalled out at one to two percent annually. AI patents exploded three hundred percent between 2020 and 2025, but follow-up analysis shows only ten percent of the new models offered truly substantial improvements. It suggests we're getting better and better at doing the same things, which is the classic sign of a technology wave reaching maturity, not one that's just beginning.

Finally, as if to prove how complex this all is, Lex Fridman just released a four-plus-hour conversation with Sebastian Raschka and Nathan Lambert. It covers everything from AGI timelines to scaling laws to the future of compute infrastructure. It's a snapshot of a field so sprawling and fast-moving that it now takes a marathon session just to map the key debates. The era of simple breakthroughs is over. We are now in the era of complex consequences. Let’s go back to that OpenAI memo. Because it’s not just about one company. It’s about the architecture of the entire AI economy. For years, the story was simple: OpenAI builds the world's best models, and Microsoft provides the world's best distribution.

Microsoft invested thirteen billion dollars into OpenAI, giving it the Azure cloud credits it needed to train massive models like GPT-5. In return, Microsoft got exclusive rights and deep integrations into its own products, from Office to Bing to its enterprise sales channels. It looked like a perfect symbiosis. But the memo, which was cited by The Meridiem, reveals what was happening behind the scenes. It claims that Microsoft’s control over enterprise sales became a bottleneck. OpenAI felt that its "ability to reach enterprise clients" was being limited. In other words, Microsoft wasn’t just a partner; it was a gatekeeper.

Every big company that wanted to use OpenAI's technology had to go through Microsoft's sales team, and presumably, Microsoft's pricing structure. So OpenAI just made a move for independence. They are now partnering with Amazon Web Services. This isn't just about being "multi-cloud." This is about building a completely independent distribution channel. It’s a direct challenge to Microsoft's control. OpenAI is saying, "We built the product that everyone wants. We should be the ones who own the relationship with the customer." They are moving to own their own toll booth, not just collect a fee from Microsoft's.

Some analysts are cautious, noting that this adds complexity and that Microsoft's deep integrations are still a huge advantage. But that misses the point. This is a power struggle. OpenAI is betting that its brand and its technology are strong enough to pull enterprise customers away from the cozy Microsoft ecosystem and directly into its own orbit, via AWS. It's a declaration that the most valuable asset in AI isn't the cloud infrastructure. It's the model itself. And this is where the conversation gets much bigger. This fight for control... what are they fighting for, exactly? This brings us to that Hacker News thread.

The central idea comes from Carlota Perez, an economist who studies technological revolutions. She argues they come in two phases. First is the "installation phase," where a new technology—like steam, or electricity, or the microchip—is installed into the economy. This phase is characterized by financial speculation, infrastructure build-out, and market saturation. It feels like a frenzy. But a lot of the value is just about replacing old systems with new, more efficient ones. Then comes the "deployment phase." This is when the technology becomes so widespread and understood that it starts to create entirely new industries, new social structures, and new ways of life.

This is where the real, broad-based productivity boom happens. The unsettling question being asked this week is this: What if AI isn't the start of a new deployment phase? What if it's the final, most powerful tool of the installation phase of the Information Age? Think about it. For three decades, the digital revolution has been about moving information around faster. The internet, the cloud, mobile computing... they were all about that. AI, in its current form, is the ultimate optimization engine for that world. It can write code faster, summarize documents better, and manage logistics more efficiently.

It is making the existing digital economy run at peak performance. But is it creating a new one? The numbers suggest... maybe not. That plateau in productivity gains at one to two percent is a huge warning sign. If AI were a true revolution on the scale of the internet, we would expect to see that number climbing. Instead, it’s flat. The fact that investment peaked at two hundred billion dollars in 2025, and now the focus is on who captures that market, also fits the pattern. The gold rush is over, and now the players are fighting over the established territory. This reframes the entire AI landscape. The race to build bigger and bigger models, the fight between OpenAI and Microsoft, the consolidation of power in Big Tech...

it all looks different. It's not necessarily a race to build the future. It could be a race to become the ultimate landlord of the present. To own the most efficient, most optimized version of the world we already have. Of course, about a quarter of the commenters on that thread push back. They argue that AI could still unlock massive gains in entirely new sectors, like drug discovery or materials science, delivering fifty percent or more in efficiency. They believe the deployment phase is still coming. And they might be right. But the conversation has shifted. For the first time, there is a credible, data-backed argument that the AI boom might be a story of consolidation, not creation.

That it's accelerating the end of the digital wave, not starting a new one. So this week, we saw the battle lines being drawn more clearly than ever. OpenAI is making a break for strategic independence, trying to control its own destiny and capture the value it creates. DeepMind is hiring philosophers to grapple with the profound ethical endgames of this technology. And regulators are waking up to the fact that a handful of companies are building insurmountable advantages in compute and data. All of this is happening at the exact moment a serious debate has begun about the true economic nature of this revolution.

The key players are fighting for control of the kingdom, building walls and moats. But for the first time, people are asking if the kingdom itself is expanding, or if they're just fighting over the last, richest plots of land. The architecture of power in AI is being finalized. The question this week raised is what that power will be used to build — a new economic world, or just a more perfectly managed version of the old one.