The Amped Journal

Palantir CEO Alex Karp used a Wednesday CNBC "Squawk Box" interview, booked to cover Palantir's expanded Nvidia partnership, to launch a nearly 20-minute broadside against the AI industry's token-based pricing. "Something has gone completely wrong," he said, calling per-token billing a "wealth tax" that leaves enterprises paying for tokens that create no value while handing over their intellectual property.

The rant followed Palantir's own nine-point "AI sovereignty" manifesto, posted on X the day before, telling companies and governments to keep their data and model weights in-house rather than outsource them to labs like OpenAI and Anthropic. It called "tokenmaxxing" - chasing high token usage - "the addictive feeling of false progress." Karp framed the Nvidia tie-up, which brings Nvidia's Nemotron models into Palantir's Sovereign AI platform, as the alternative: customers "own the means of production" instead of transferring it to a third party.

When anchor Becky Quick told him "you sound pretty angry," Karp reframed it as "the voice of American business," daring viewers to privately call any CEO and hear they're "twice as livid" as he appeared. The clip went viral on X - some praised him for calling out AI hype, others pushed back over Palantir's defense ties and Karp's pro-Israel remarks elsewhere in the interview.

Reuters reports OpenAI has discussed giving the US government a 5% stake in the company. The reported talks come as AI firms face growing scrutiny in Washington over potential misuse of advanced models and questions about whether Americans will share in the sector's profits.

No further details on structure or timeline were disclosed in the report.

GV Managing Partner Dave Munichiello discussed the firm's portfolio company Modular and its acquisition by Qualcomm in an interview with Crunchbase News, according to a post from GV. Munichiello framed the deal as an example of how GV partners with founders to build through hard moments.

Modular builds AI infrastructure and compiler tooling. Terms of the acquisition were not disclosed in the post.

The Trump administration is finalizing voluntary safety standards for powerful AI models with advanced cyber capabilities, including benchmarks co-designed with Anthropic, release timelines, and access rules for US and foreign users. The framework stems from a June 2 executive order and involves agencies including the NSA and CAISI sharing classified benchmarks to protect infrastructure while promoting US AI leadership.

It follows recent interventions including limits on OpenAI's GPT-5.6 preview to approved partners and Anthropic's own pause of its Fable 5 and Mythos models over jailbreak risks. Cursor confirmed Fable 5 is available again in its product, noting the model leads all models on CursorBench but is the most expensive per task to run. An announcement on the broader standards could come next week.

Bloomberg reports China's Kling AI has raised an initial $2 billion in venture capital funding. Kling is a spinoff from Kuaishou Technology and is using the funding to expand its video AI operations.

Microsoft announced Microsoft Frontier Company, a new operating business unit dedicated to delivering what it calls "Frontier Transformation" to customers worldwide. The unit commits $2.5 billion and 6,000 employees, including industry experts, change management specialists, and AI engineers, who will embed within client organizations to co-design, deploy, and improve AI systems.

The unit is led by Rodrigo Kede Lima.

The Associated Press reports Russia launched a large-scale attack on Ukraine's capital overnight into Thursday, killing at least eight people and injuring dozens. Loud explosions shook Kyiv for hours during the assault.