This week, money spoke louder than benchmarks.
I’ve filtered out the noise for you
1. SpaceX will buy Cursor for $60B (or pay $10B for trying)
SpaceX announced an option to acquire Cursor for $60B this year or pay $10B to collaborate. Cursor was negotiating a round at $50B with its CEO Michael Truell (25 years old, net worth $1.3B). The deal precedes SpaceX’s IPO with a target valuation of $1.75T.
Why it matters: In a market where all models converge on capabilities, the real fight has shifted to whoever owns distribution.
2. Google invests up to $40B in Anthropic, its own rival
Google will invest up to $40B in Anthropic: $10B immediately at a $350B valuation plus $30B tied to milestones. This adds to Amazon’s $25B. Anthropic’s annualized revenue now exceeds $30B vs. $9B at the end of 2025.
Why it matters: Google has Gemini, TPUs, and cloud, but is investing in its dominant competitor in enterprise where Claude Code leads.
3. OpenAI launches GPT-5.5: the model for real work, not for chatting
OpenAI launched GPT-5.5 for Plus/Pro/Business/Enterprise with an updated version of Codex. Better results, same latency, lower consumption. Greg Brockman describes it as a step toward agentic computing.
Why it matters: It’s the first time OpenAI removes chat from the center of a launch.
For sales teams tired of cold leads, slow customer responses, and manual processes, Dapta is the ultimate tool.
Dapta is the leading platform for creating AI sales agents specifically designed to increase inbound lead conversion. Respond to your leads in less than a minute with voice AI and WhatsApp that converts.
If you want your team to sell more while AI handles the complex stuff, you have to try it.
4. Kimi K2.6: open source agents coordinating 300 sub-agents in parallel
Moonshot AI released Kimi K2.6 (1T parameters, 256K context). It scales its Agent Swarm to 300 sub-agents with 4,000 simultaneous steps vs. 100/1,500 in K2.5. It beats Claude Opus 4.6 and GPT-5.4 on Humanity’s Last Exam.
Why it matters: First open source model documenting autonomous tasks of 12+ hours with thousands of tool calls, at $0.74 per million input tokens.
5. DeepSeek V4: frontier quality at one-sixth the price
DeepSeek released V4 (1.6T parameters, 1M token context) and V4-Flash (284B). V4-Pro hits 80.6% on SWE-Bench Verified (0.2 points below Claude Opus 4.6). Its output pricing is one-sixth that of Opus 4.7 and GPT-5.5.
Why it matters: Second time DeepSeek shifts the industry with an open model closing in on the closed leaders.
Bonus: Content of the Week
I’ve been a couple of weeks in San Francisco with the Dapta team and the things we’ve seen and experienced are impressive. The best companies in the world are putting out incredible advances, but what does that mean for us regular entrepreneurs? It means opportunities! We’re documenting the adventures in San Francisco on our Youtube.
