/
Navigation
Chronicles
Browse all articles
Explore
Semantic exploration
Research
Entity momentum
Nexus
Correlations & relationships
Story Arc
Topic evolution
Drift Map
Semantic trajectory animation
Posts
Analysis & commentary
Pulse API
Tech news intelligence API
Browse
Entities
Companies, people, products, technologies
Domains
Browse by publication source
Handles
Browse by social media handle
Detection
Concept Search
Semantic similarity search
High Impact Stories
Top coverage by position
Sentiment Analysis
Positive/negative coverage
Anomaly Detection
Unusual coverage patterns
Analysis
Rivalry Report
Compare two entities head-to-head
Semantic Pivots
Narrative discontinuities
Crisis Response
Event recovery patterns
Connected
Search: /
Command: ⌘K
Embeddings: large
TEXXR

Chronicles

The story behind the story

← → days · ↑ ↓ browse · Enter similar · o open

Google debuts Code as Policies, a tool that uses AI models to generate code for robots to perform a variety of tasks, under an open-source license on GitHub

Google LLC today debuted an internally developed software tool that can reduce the amount of effort involved in training a robot to perform new tasks.

SiliconANGLE Maria Deutscher

Context & Ripple Effects

Code as Policies extends a habit Google established years ago when DeepMind released its entire AI training platform on GitHub: ship frontier AI tooling as open source to seed an ecosystem around it. The difference here is the output — instead of giving researchers a training platform, Google is handing developers a generator that turns model outputs into executable robot behavior.

The move also reads as an early marker in Google's broader shift toward AI-written code. The company has since reported that 75% of new code created inside Google is generated by AI and reviewed by human engineers, up from half the prior fall, while shipping agent tooling like Jules, the autonomous bug-fixing assistant built on Gemini 2.0. Code as Policies applies that same 'model writes the program' idea to physical machines.

First-order effects

  • Robotics developers get a free, inspectable alternative to hand-writing task-specific control code, lowering the effort of teaching robots new tasks — and every improvement lands on GitHub rather than inside a proprietary stack.
  • Google positions itself as the default starting point for language-model-driven robot programming, the same ecosystem-seeding role its DeepMind platform played for researchers.

Second-order effects

  • Rivals building robot software stacks face pressure to match an open-source baseline: charging for basic natural-language-to-behavior generation gets harder when Google gives it away.
  • GitHub gains another high-profile repository that funnels robotics work through its platform, reinforcing its pull as the distribution layer for Google's AI tooling alongside its own Copilot agent features.

Third-order effects

  • If the pattern holds, robot programming migrates from specialist engineers writing control code toward operators specifying tasks in natural language and letting models emit the policy — collapsing a labor category into prompt-and-review work, mirroring what Google reports happening to its own internal codebase.
  • Open-source release shifts the battleground from who owns the generator to who controls the surrounding stack — simulation, safety validation, and fleet management become where vendors differentiate once the core code-generation layer is commoditized.

The trend: Google is moving from open-sourcing AI research platforms to open-sourcing tools that let AI models write operational code directly — for software agents first, and now for robots.

Discussion

  • @jackyliang42 Jacky Liang on x
    How can robots perform a wide variety of novel tasks from natural language? Execited to present Code as Policies - using language models to directly write robot policy code from language instructions. See paper, colabs, blog, and demos at https://code-as-policies.github.io/ long …
  • @haydenfield Hayden Field on x
    In Google AI's demo room, one exhibit shows “robots that write their own code,” noting that large language models can generate “robot code” from natural language. https://twitter.com/...
  • @googleai @googleai on x
    PaLM, our latest & most capable language model, can unlock new possibilities in robotics. For example, read about Code as Policies, an approach that, when given natural language instructions, uses language models to write code that controls robots→ https://ai.googleblog.com/... (…