SpaceX's valuation jumped another trillion since Friday's IPO and briefly passed Amazon at $2.6T. Leaked audited accounts show OpenAI losing billions of dollars a year as revenue is dwarfed by R&D and operating expense. GLM-5.2 took the lead spot on Artificial Analysis's open-weights index, validating the sovereign-and-open thread that started with India and the Netherlands. The Fable saga keeps spinning: Anthropic's business is reportedly growing through the fight, the company paused the token-based Agent SDK billing that would have hit power users hardest, and Wired argues 'dangerous models are coming no matter what.' The DOJ told a court xAI's unpermitted gas turbines are a matter of 'national, economic, and energy security' — Grok is now formally on the side of the Pentagon. A WordPress VIP survey found 60% of US consumers consider 'AI' in brand messaging a turnoff, and Robinhood walked back its layoff-blaming-AI framing under pressure.
SpaceX gained $1T of valuation in three trading days and briefly passed Amazon at $2.6T. Leaked OpenAI financials show audited revenue growing fast but dwarfed by R&D and operating expense — multi-billion annual losses. The two ends of the AI capital story land in the same news cycle.
SpaceX added a trillion dollars of valuation in three trading days and briefly passed Amazon at $2.6T.
Audited OpenAI financials show revenue growing fast but dwarfed by R&D and operating expense — multi-billion-dollar annual losses.
The Verge confirms the SpaceX-Cursor $60B acquisition — official today.
Ars frames the deal: separately, neither SpaceX-AI nor Cursor could compete; together they hope to.
GLM-5.2 took the top spot on Artificial Analysis's open-weights index, validating a month of compounding Chinese, European, and emerging-market releases. The Netherlands officially announced GPT-NL as a sovereign LM. Qwen-Robot Suite added physical-AI foundation models to the open frontier. Anthropic published a 'Founder's Playbook' on building AI-native startups — bidding for the developer mindshare local-AI sentiment is contesting.
GLM-5.2 takes the top spot on Artificial Analysis's open-weights intelligence index (154 HN points).
Zhipu's official GLM-5.2 launch blog — long-horizon agent focus.
TNO announces GPT-NL — a Dutch sovereign LM project (222 HN points).
Alibaba ships Qwen-Robot Suite — physical-AI foundation models for embodied agents (184 HN points).
Anthropic publishes a founder's playbook for AI-native startups (106 HN points).
Anthropic's business is reportedly growing through the fight with the White House; the company paused token-based Agent SDK billing that would have hit power users hardest; Wired argues 'dangerous models are coming no matter what'; Stratechery weighs in on the jailbreak problem and the SpaceX-Cursor deal in the same column.
TC: Anthropic's business-user growth is accelerating through the Fable fight — the controversy is functioning as a marketing event.
Anthropic paused a Monday-planned token-based Agent SDK billing change that would have heavily increased power-user costs.
Stratechery argues the administration is wrong about Fable but the responsibility ultimately falls on Anthropic to navigate it.
Wired argues the Fable shutdown hides a glaring truth — frontier capability proliferation is structural, not vendor-controllable.
The DOJ formally argued in court that xAI's unpermitted gas turbines are a matter of 'national, economic, and energy security' — Grok is now publicly positioned as Pentagon-aligned. The Pentagon boasts of using AI to write Congress-mandated reports. Meanwhile, a WordPress VIP survey says 60% of US consumers find 'AI' in brand messaging a turnoff, and Robinhood walked back its layoff-blaming-AI framing.
DOJ argues the Pentagon needs xAI to keep using its unpermitted gas turbines — Grok is now formally on the security-protected list.
Ars on the administration trying to dismiss NAACP's Clean Air Act suit on national-security grounds.
Pentagon claims 1.5M personnel are using generative AI tools, including for Congress-mandated reports.
WordPress VIP survey: 60% of US consumers say 'AI' in brand messaging is a turnoff.
Robinhood walked back the 'restructuring around AI' framing for its 10% layoffs after pushback.
OPD-Evolver applies on-policy distillation to evolve holistic agent capability. MotionVLA targets humanoid motion control. ACE-Ego-0 unifies egocentric human and robotic data for VLA pretraining. ProCUA-SFT advances computer-use agents. The research wave keeps pushing past flat policy training toward composed, embodied skills.
On-policy distillation that evolves holistic agent capability — past per-task fine-tuning.
VLA model targeting humanoid motion control — language-conditioned whole-body movement.
Unifies egocentric human video and robotic teleoperation data for VLA pretraining — exploits both sources.
Computer-use-agent SFT recipe — practical advance for production CUAs.