🧬 AI Lab Updates
OpenAI Releases GPT Rosalind — First Biology-Specific Large Model
- Source: OpenAI Official
- OpenAI launches GPT Rosalind, a biology-domain model named after DNA pioneer Rosalind Franklin
- Focuses on protein structure prediction, genome analysis, drug discovery
- Marks OpenAI’s strategic expansion from general AGI to vertical scientific domains
Comment: Great naming choice. Rosalind Franklin was a key figure in DNA structure discovery who was long overlooked. OpenAI honoring her name while launching a bio model sends a strong brand message. AI for Science is now an official OpenAI battleground.
Anthropic Launches Anthropic Science Initiative
- Source: Anthropic Research
- Anthropic establishes a dedicated science program, applying Claude to fundamental research
- Covers mathematical proofs, physics simulations, chemical discovery
- Echoes GPT Rosalind — two major AI labs betting on scientific AI simultaneously
Comment: This isn’t a coincidence. Both OpenAI and Anthropic announcing science AI initiatives in the same period signals industry consensus: the next value frontier isn’t chatbots, it’s AI-accelerated scientific discovery. Whoever achieves reproducible scientific breakthroughs first controls the next narrative.
Anthropic Releases Long-running Claude — Persistent Sessions
- Source: Anthropic Research
- Claude can now run complex tasks for extended periods without frequent human intervention
- Supports multi-step workflows, background processing, automatic recovery
- Combined with Automated Alignment Researchers, forming a self-improvement loop
Comment: Long-running sessions are foundational infrastructure for Agent use cases. Claude is taking a key step from “conversational AI” to “autonomous execution AI.”
Anthropic Introduces Claude Think Tool
- Source: Anthropic Engineering
- New structured reasoning tool that makes Claude explicitly show its thinking process on complex tasks
- Supports safe code execution under MCP protocol
Comment: The Think Tool fundamentally addresses Agent interpretability. Making the model “speak its thoughts” is crucial for debugging and building trust.
🤖 Products & Applications
OpenAI Codex Evolving into a Super App
- Source: The Rundown AI
- Codex is no longer just a coding tool — integrating agents, browser automation, file processing
- OpenAI’s ambition is to make Codex the “operating system” of the AI era
- One entry point for programming, research, and automation
Comment: Codex’s evolution from code assistant to super app is heating up competition with Cursor and Windsurf. OpenAI’s advantage lies in deep model-level integration, which third-party editors can’t easily replicate.
Claude Opus 4.7 Released
- Source: TLDR AI
- Anthropic releases Opus 4.7, continuing iteration of its flagship model
- Perplexity launches Personal Computer — AI search enters a new phase
Comment: Opus series iteration speed is accelerating. The GPT-5 vs Opus 4 competition ultimately benefits users — model quality is rapidly converging while prices trend downward.
📊 Industry Analysis
Stratechery: Servers, Satellites, and Stars
- Source: Stratechery
- Ben Thompson analyzes AI infrastructure costs, Amazon-Globalstar satellite partnership
- Core thesis: AI costs are shifting from training to inference; infrastructure ROI becomes the key question
Comment: AI economics is shifting from “who can train the largest model” to “who can run inference cheapest.” This transformation has profound implications for the entire industry landscape.
Pragmatic Engineer: How Codex is Built
- Source: Pragmatic Engineer
- Gergely Orosz provides a deep dive into OpenAI’s Codex infrastructure
- Covers sandbox environments, code execution, security isolation
Comment: Codex’s engineering architecture is worth studying for every team building AI products. Security sandbox + model inference + user experience — all three are indispensable.
Simon Willison: Datasette 1.0 Progress
- Source: Simon Willison
- Datasette advances toward 1.0 release as Simon continues updating this data exploration tool
- With AI integration, Datasette is becoming a bridge between data analysis and LLM applications
Comment: Simon is one of the most noteworthy independent developers in the LLM practice space. Datasette’s AI integration approach offers valuable lessons for small teams.
🌐 Other
Google AI: Smart Summer Travel Suggestions
- Source: Google AI Blog
- Google Search AI mode adds travel planning features
- Integrates Maps, flights, hotel data for personalized itinerary suggestions
One-Sentence Summary
This week’s AI industry central theme is “AI for Science” — OpenAI and Anthropic both announced scientific AI initiatives, Codex evolves from a coding tool into a super app, and Long-running Claude makes autonomous agents a reality. AI is transitioning from a “conversation tool” to a “research partner and autonomous executor.”