Good morning. It's Thursday, October 8, and we're covering how AI is streamlining government services, Microsoft's push to bring AI onto PCs, and Danaher's plans for autonomous research labs.

Plus: yesterday's Real vs AI verdict, a new poll on AI’s role in government, and a guest essay on why your AI strategy should look more like R&D.

Life sciences company Danaher plans to open an AI-powered autonomous research lab in early 2027. The company says a combination of laboratory automation, connected instruments and AI could move some research programs from question to validated answer up to eight times faster.

Elon Musk says SpaceX’s Grok Bot will route some tasks to rival systems, including Anthropic’s Claude, Midjourney and Suno. That makes it an orchestrator of specialized models rather than a single-model assistant; the next test is whether those handoffs work reliably. (Paywall)

OpenAI introduced Intelligent UI, a GPT-6 feature that can combine text with charts, forms, buttons and interactive tools inside ChatGPT. It is rolling out first to paid plans, with Free and Go access following next, so the question is whether people find generated interfaces more useful than a standard replies.

Anthropic released Claude Haiku 5.5, which it calls its fastest and most capable small model. It is designed for high-volume jobs such as summarization, browser use and subagents, making price, latency and reliability the measures to watch against larger models.

Former New Jersey chief AI strategist Beth Simone Noveck argues that governments can use AI to make public services faster and easier to access. She points to New Jersey matching records to reach more children with summer food benefits, California using AI to analyze wildfire-survivor feedback, and Britain sorting tens of thousands of public comments in hours instead of weeks.

The useful question is not whether software should replace public officials, but where it can remove paperwork without hiding judgment or accountability. Watch whether governments pair these systems with trained staff, multiple model options and clear checks for mistakes before AI moves from processing information to shaping decisions. → Read the full article here. (Paywall)

AI can help governments process records, analyze public feedback and deliver services faster—but deciding policy on behalf of citizens is a much bigger step.

In 2023, Duolingo began building new AI features on top of GPT-4. The company didn't build the model. It built everything around it: the scenarios, curriculum and product experience that turn a general-purpose model into a language tutor, plus Birdbrain, its own system for judging what each learner can handle next.

Morgan Stanley did the same, putting GPT-4 inside an internal assistant that searches the firm's own research for its advisors. In both cases the model was the commodity and the system around it was not. → Read the full article here.

Microsoft and NVIDIA unveiled a new generation of Windows systems built around RTX Spark hardware and on-device AI. Microsoft also introduced hybrid intelligence, which can split work between local hardware and cloud models so an application can keep sensitive data nearby while calling on more computing power when needed.

Local AI could make agents faster, more private and less dependent on an internet connection, but developers still need to prove that people want these features on everyday PCs. Watch the experimental Copilot and developer-tool previews later this month, along with whether software makers build useful workflows for the new hardware. → Read the full article here. (Paywall)

Parallel Systems, founded by former SpaceX engineers, raised $100 million in Series C funding to scale battery-powered autonomous freight vehicles. Its Panther vehicles can carry several tons of freight for up to 500 miles, either independently or in shorter, uncoupled platoons.

The company is targeting short-haul freight, where railroads have ceded much of the market to trucking as they increasingly favor longer trains and fixed schedules. Parallel has already received Federal Railroad Administration approval and has tested its system across 160 miles of track near the Port of Savannah, Georgia. → Read the full article here.

Google Opens a Prompt-Based Game Builder: Playground lets U.S. adults create, play and share browser games from text prompts.

Executives Target GPU Crunch: Anjney Midha and former tech executives launched a company to make AI compute cheaper for startups. (Paywall)

Meta Targets Dangerous Ads With AI: New tools detect ads that secretly direct users to child sexual abuse material hosted off-platform.

Sriram Krishnan Seeks $500M AI Fund: The former White House adviser is raising $500M for growth-stage AI companies tied to national security.

Google Opens SynthID Detection: Anyone can check images, video and audio for Google and participating AI companies’ watermarks.

Yesterday's image was REAL. Here's how you voted: 54% thought it was AI, while 46% correctly identified it as real.

Thanks for reading. See you next time!

— Matthew Berman, Nick Wentz & the Forward Future Team