OpenAI Officially Retires GPT-4o, GPT-4.1 and Other Older Models: GPT-5.2 Becomes ChatGPT Default L1
Confidence: High
Key Points: OpenAI officially retired GPT-4o, GPT-4.1, GPT-4.1 mini, and o4-mini models from ChatGPT on February 13. GPT-5.2 becomes the default model for all users. This move comes as the vast majority of usage has shifted to GPT-5.2, with only 0.1% of users per day still choosing GPT-4o.
Impact: ChatGPT Business, Enterprise, and Edu customers can retain GPT-4o access in custom GPTs until April 3. API users are unaffected, as these models remain available. The retirement has triggered strong backlash from some users, particularly those who used 4o for companion interactions.
Detailed Analysis
Trade-offs
Pros:
GPT-5.2 offers stronger performance
Simplifies product line maintenance
Also increases attachment limit to 20 files
Cons:
4o loyal users lose their preferred option
5.2 has stricter safety guardrails
Quick Start (5-15 minutes)
All ChatGPT users automatically upgrade to GPT-5.2
API users continue using 4o series without changes
Enterprise customers evaluate migration plans before April
Recommendation
Enterprise users relying on older models should complete migration testing before April 3. Evaluate whether GPT-5.2 meets existing workflow requirements.
GPT-5.2 Achieves New Discovery in Theoretical Physics: Challenges 40-Year Gluon Amplitude Assumption L1
Confidence: High
Key Points: OpenAI, in collaboration with Princeton's Institute for Advanced Study, Harvard, Cambridge and other universities, published a preprint showcasing gluon amplitude formulas proposed by GPT-5.2. The research challenges the decades-old textbook assumption that 'single negative helicity gluon tree-level amplitudes are zero,' discovering non-zero amplitudes can be generated under 'semi-collinear region' conditions.
Impact: GPT-5.2 Pro proved its own proposed formula (Equation 39 in the paper) after 12 hours of sustained thinking. This discovery has been extended to the graviton field. UC Santa Barbara physics professor Nathaniel Craig called it 'journal-quality research advancing the frontiers of theoretical physics,' demonstrating the future of AI-assisted science.
Detailed Analysis
Trade-offs
Pros:
AI's first verifiable new discovery in theoretical physics
New paradigm for human-AI collaborative research
Potentially opens more cross-disciplinary applications
Cons:
Still requires human verification and proof
May trigger discussions on academic attribution
Quick Start (5-15 minutes)
Read the arXiv preprint for technical details
Physics researchers can explore related extension directions
Recommendation
Scientific research institutions should follow the latest developments in AI-assisted research and evaluate the possibility of integrating large language models into research workflows.
OpenAI Introduces Lockdown Mode and Elevated Risk Labels: Defending Against Prompt Injection Attacks L1
Confidence: High
Key Points: OpenAI released two new security features: Lockdown Mode, an optional advanced setting for users with high security needs, and Elevated Risk labels to mark features with additional security risks in ChatGPT, Atlas, and Codex.
Impact: Lockdown Mode restricts web browsing to cached content only, preventing sensitive data leakage through browsing. Currently available only for ChatGPT Enterprise, Edu, Healthcare, and Teachers editions, with consumer versions coming in the next few months.
Detailed Analysis
Trade-offs
Pros:
Effectively defends against prompt injection attacks
Suitable for high-risk users (such as executives, security teams)
Unified risk labeling improves transparency
Cons:
Some features disabled in Lockdown Mode
Currently limited to enterprise versions
Quick Start (5-15 minutes)
Enterprise admins enable Lockdown Mode in ChatGPT settings
Follow usage guidelines for features with Elevated Risk labels
Recommendation
Enterprise users handling sensitive data should evaluate enabling Lockdown Mode. Security teams need to understand the risks of Elevated Risk features.
Anthropic Partners with CodePath: Claude Enters America's Largest University Computer Science Program L1
Confidence: High
Key Points: Anthropic partnered with nonprofit education organization CodePath to integrate Claude into America's largest university computer science program. This collaboration will help college students use AI tools during their learning process.
Impact: CodePath serves tens of thousands of college students, and this partnership will significantly expand Claude's influence in the education sector while cultivating AI usage habits among the next generation of developers.
Detailed Analysis
Trade-offs
Pros:
Expands Claude's education market coverage
Cultivates future user base
Promotes AI education popularization
Cons:
Educational scenarios require special attention to academic integrity
Quick Start (5-15 minutes)
CodePath students can access Claude through the program
Educational institutions can reference this partnership model
Recommendation
Educational institutions can follow Anthropic's education partnership programs and evaluate the possibility of similar integrations.
Godot 4.7 dev 1 Development Snapshot Released: VirtualJoystick, Vulkan Ray Tracing Foundation L2GameDev - Code/CI
Confidence: High
Key Points: Godot 4.7's first development snapshot released, with 127 contributors submitting 311 improvements. New features include built-in VirtualJoystick node, DrawableTexture, Path3D collider snapping, Vulkan ray tracing infrastructure, and Windows HDR display prototype support.
Impact: Simplifies touch input development for mobile game developers, Vulkan ray tracing lays foundation for future visual upgrades.
Detailed Analysis
Trade-offs
Pros:
Simplifies mobile development
Approaches commercial engine capabilities
Cons:
Development snapshots not suitable for production environments
Key Points: ChatGPT now allows users to attach up to 20 files in a single message (previously 10), facilitating document set analysis, multiple file comparison, or providing broader context.
Impact: Improves efficiency in handling multi-document tasks.
Detailed Analysis
Trade-offs
Pros:
More efficient document analysis
Cons:
May increase token consumption
Quick Start (5-15 minutes)
Upload multiple files directly in ChatGPT
Recommendation
Users who need to process large volumes of documents can make good use of this feature.