Get the latest ideas from The AI Daily Brief: Artificial Intelligence News and Analysis.
Plus the best new takeaways about artificial intelligence from other top podcasts — read in minutes, not hours.
or
By continuing, you agree to Podbrain's Terms and Privacy Policy.
The AI Daily Brief covers major developments in AI consulting, new interaction paradigms, and regulatory updates. Host Nathan Lambert discusses OpenAI's official launch of DeployCo, a $4 billion consulting venture, and Anthropic's crackdown on unauthorized stock trading.
The main focus examines Thinking Machines' breakthrough 'interaction models' - a new AI architecture that processes continuous micro-turns rather than traditional turn-based exchanges. Former OpenAI CTO Mira Murati's company demonstrates real-time visual recognition, simultaneous translation, and background task management.
Additional coverage includes the White House backing away from FDA-style AI regulation after industry concerns, and President Trump assembling a tech delegation for China trade talks that notably excludes NVIDIA's Jensen Huang.
OpenAI's $4 Billion DeployCo Consulting Launch
OpenAI officially launched the OpenAI Deployment Company (DeployCo) as a separate consulting entity with $4 billion in funding at a $10 billion pre-money valuation, led by TPG with Advent International, Bain Capital, and Brookfield as co-founding partners.
The venture operates as a 'buy-in cost to skipping the line' for AI transformation services, with investors gaining first access to engineers for their portfolio companies.
DeployCo acquired engineering firm Tomorrow to gain 150 experienced AI deployment staff immediately, with expectations of more M&A to meet massive demand.
Secondary Market Crackdown Triggers Price Crashes
Anthropic updated support documentation declaring all SPV transfers void and specifically naming firms offering unauthorized stock access, triggering a 50% price crash in gray markets.
OpenAI issued similar warnings about unauthorized transfers being 'legally void without approval,' with lawyer Gabriel Shapiro noting potential for 'an avalanche of lawsuits.'
Casey Craig described the layered complexity: 'You are four layers of financial abstraction and broker crime away from touching an actual Anthropic share certificate.'
The crackdown highlights broader issues with private market structures where companies stay private longer, blocking retail investor participation in startup upside.
White House Retreats from FDA-Style AI Regulation
National Economic Council Chairman Kevin Hassett walked back his FDA comparison after massive industry reaction, stating 'nobody has an idea that we should do something like bring in a giant new bureaucracy to approve AIs.'
The current approach involves administration officials working directly with AI labs to ensure models 'aren't going to cause an extreme amount of harm' before public release.
Hassett acknowledged the communication misstep: 'I probably shouldn't have called it the FDA.'
Thinking Machines Unveils Continuous Interaction Models
Former OpenAI CTO Mira Murati's Thinking Machines introduced interaction models that process 200-millisecond micro-turns instead of traditional turn-based exchanges, creating 'constant two-way exchange with the user.'
The architecture combines a real-time interaction model with a background model for longer reasoning, allowing simultaneous conversation and task execution.
Demonstrated capabilities include simultaneous translation while speakers are still talking, visual recognition of new people entering frame, and proactive reminders based on visual cues like posture changes.
TML created new benchmarks TimeSpeak and QSpeak to measure proactive audio capabilities, as existing benchmarks couldn't capture the new interaction paradigm.
The GUI Moment for AI Interaction
TML researcher Claire Birch argued that current AI chat interfaces are 'surprisingly CLI-like' and reward 'verbal fluency, abstraction, and procedural skill' rather than natural human collaboration patterns.
The vision positions interaction models as 'the GUI moment' where users 'no longer have to think like the computer, or like the AI, or like the prompt engineer' to access capabilities.
The approach addresses what TML calls the 'collaboration bottleneck' where current models 'experience reality in a single thread' and freeze perception until generation completes.
From The AI Daily Brief: Artificial Intelligence News and Analysis. Get a note like this from every new episode.