July 29, 2026
Encore AI just raised $30 million to do something specific: mine customer calls, messages, and CRM records to figure out which sales techniques actually close deals, then bake those patterns into playbooks for AI agents, according to TechCrunch. That's a narrower bet than the usual 'AI copilot for everything' pitch, and I think that's exactly why it's worth watching. Most sales AI so far has been about summarizing calls or drafting follow-up emails. Encore is trying to reverse-engineer what your best rep does differently from your average rep, then teach an agent to do the same thing at scale. If it works, it's a genuine threat to the idea that sales coaching requires a human sales manager sitting in on calls. If it doesn't, you've handed a vendor a lot of sensitive conversation data for a playbook that might just restate what a good manager already knew. Either way, this is the kind of thing that only works if the underlying CRM data feeding it is clean and centralized -- teams that already have call, message, and deal data scattered across four disconnected tools are going to get garbage playbooks no matter how good Encore's model is.
The operator of the largest power grid in the US is now considering temporary power cuts to data centers to avoid blackouts, TechCrunch reported. This is the moment the AI buildout stops being an abstract capex story and becomes a literal supply constraint. Every model release, every 'we're scaling inference' announcement, every new data center groundbreaking has been implicitly betting that power would just be there. Now a grid operator is saying, in effect, that it might not be -- at least not at the pace the industry wants. For a business evaluating AI vendors, this matters less as a moral story and more as a reliability question: if your AI-dependent workflows depend on compute that a grid operator can throttle on short notice, that's a new kind of vendor risk that wasn't on anyone's checklist a year ago. We flagged this trajectory before -- the grid is sending AI a bill it can't dodge -- and this week's news is that bill arriving faster than expected. I don't think this stops the buildout. But it should make buyers ask vendors harder questions about where their compute actually lives and what happens during a curtailment event.
Cursor is expanding aggressively in India with localized pricing, local hiring, and enterprise sales, calling it now the company's third-largest market, per TechCrunch -- notably timed ahead of a reported SpaceX acquisition. Meanwhile, Anthropic's Dario Amodei clarified that he doesn't oppose open-weight models on principle, but he's specifically worried about Chinese AI capability closing the gap, TechCrunch reported. Put these two next to each other and you get a clearer picture of where the real AI competition is happening: not just model quality, but geography and openness. Cursor chasing India isn't charity -- it's a bet that the next wave of paying developers isn't concentrated in San Francisco. Amodei's comments are a reminder that the loudest US AI labs are increasingly reasoning about their product decisions through a geopolitical lens, not just a technical one. I find Amodei's framing reasonable but incomplete: fearing Chinese open-weight progress is a legitimate strategic concern, but it's also a convenient argument for labs that would rather not open-source their own frontier models. Business buyers don't need to pick a side in that debate, but they should notice that pricing, availability, and even model openness are increasingly downstream of national competition, not just engineering roadmaps. If you're building on any single vendor's stack, that's one more reason to think about build vs. buy tradeoffs rather than assuming today's pricing and access will hold steady.
So which of these worries you more: an AI vendor that might get its power throttled, or one whose roadmap is shaped as much by geopolitics as by product strategy? I'd like to hear where you'd draw the line.
Sources