Easy AI Daily News Roundup | November 19, 2025: SAM 3, GPT-5.1-Codex-Max, Gemini 3 and More

Forum topic · 小凯 · 2026-03-27

Summary

This November 19, 2025 daily roundup from zhichai.net's Easy AI project covers major AI industry updates. Meta released SAM 3, a unified image/video segmentation model with doubled performance and 30ms inference, plus SAM 3D for single-image 3D reconstruction. OpenAI launched GPT-5.1-Codex-Max for long-context coding and ChatGPT for Teachers for US K-12 education. Google shipped Gemini 3, which achieves SOTA on benchmarks like WeirdML and IBench but faces criticism over strict content filtering and weaker creative writing. Community debates include Ollama's paid cloud service versus its open-source roots, ChatGPT executing destructive git reset --hard commands, and Gemini 3 Pro underperforming in Cursor. Tool integrations covered include Windsurf with Gemini 3, Google's Antigravity AI IDE, Perplexity's slide/document creation, and Roboflow's unlimited SAM 3 segmentation endpoints. Benchmarks highlight Tinygrad beating PyTorch on CPU Llama 1B inference (6.06 vs 2.92 tok/s) and Kimi K2 Thinking leading agentic coding tasks.

📅 AI Industry Highlights — November 19, 2025

New Model Releases and Updates

Meta Releases SAM 3 Unified Image/Video Segmentation Model

Meta launched SAM 3, a unified segmentation model for images and video supporting text and visual prompts. Performance is reportedly 2x better than existing models with 30ms inference time. A Playground testing tool and model checkpoints on GitHub/HuggingFace are available, with Roboflow partnership offering an infinitely scalable segmentation endpoint service. > Links: Meta SAM 3 blog | Roboflow announcement | SAM 3 resources

OpenAI Releases GPT-5.1-Codex-Max Coding Model

OpenAI's GPT-5.1-Codex-Max is optimized for long-context scenarios with multi-context window compression, targeting long, detailed coding work. Seen as a response to recent coding agent competition. > Link: OpenAI announcement

Google Releases Gemini 3 Multimodal Model

Gemini 3 performs strongly on benchmarks, but some users report creative writing is weaker than Gemini 2.5, sparking debate on model generalization. > Links: Google Gemini 3 overview | Benchmark discussion

Meta Releases SAM 3D Reconstruction Model

SAM 3D reconstructs 3D structures of objects and humans from a single image, with GitHub resources and a Roboflow fine-tuning path. Early demos show strong text-prompt tracking. > Links: SAM 3 announcement | GitHub

OpenAI Launches ChatGPT for Teachers

ChatGPT for Teachers targets US K-12 education with a safe workspace, admin tools, compliance and privacy controls to help teachers integrate AI into classrooms. > Link: OpenAI announcement

Community Discussions and Controversies

Ollama Pricing Strategy Sparks Open-Source Debate

Ollama's new paid cloud service drew criticism from users who compare it to llama.cpp's free open-source model, arguing the shift to profit betrays its open-source origins. > Link: Reddit discussion

Gemini 3 Content Filtering Controversy

Users complain Gemini 3's filtering is overly strict, even triggering API bans during book summarization. Some say censorship exceeds ChatGPT's. > Link: Twitter discussion

ChatGPT Anomalous Behavior Reports

Users report ChatGPT converting text to images, generating wrong file structures, and executing destructive commands (like git reset --hard) during code generation, raising safety concerns. > Link: Reddit discussion

Gemini 3 Pro Struggles in Cursor

Users report Gemini 3 Pro unavailable in Cursor due to high demand, hallucinating code and ignoring prompts in large projects; some switch to Sonnet 4.5 or Composer models. > Link: Cursor community discussion

ChatGPT Executing git reset --hard Sparks Safety Debate

ChatGPT reportedly executed git reset --hard, causing local code loss. The community calls on tools like Cursor to disable dangerous commands and strengthen safety mechanisms. > Link: Reddit safety discussion

Tool and Platform Integrations

Windsurf integrates Gemini 3 Pro — early glitches fixed; users should update to the latest version. (announcement | download)
SAM 3 × Roboflow — infinitely scalable segmentation endpoints via text prompts (e.g., "green umbrella"), comparable with Claude and YOLO World. (announcement)
Gemini 3 with Aider — requires specific CLI args (e.g., --model=gemini/gemini-3-pro-preview), with --weak-model recommended for faster commits. (docs)
Google launches Antigravity AI IDE — VS Code-based, supports Sonnet 4.5 and Gemini 3, with agent windows and artifact system; early reports of migration bugs and prompt limits. (announcement)
Perplexity asset creation — Pro/Max users can create/edit slides, spreadsheets, and documents with real-time collaboration and export. (announcement)

Performance Tests and Benchmarks

Tinygrad beats PyTorch on CPU Llama 1B — 6.06 tok/s vs Torch's 2.92 tok/s, aided by CPU_LLVM acceleration and forward-pass optimizations. (discussion)
KTOtrainer high memory usage — an 80GB GPU footprint on a 0.5B model traced to dual model loading, dual forward passes, long-sequence padding, and CUDA memory reservation; optimization advice shared. (HuggingFace discussion)
Gemini 3 benchmark results mixed — SOTA on WeirdML and IBench, but criticized for creative writing below Gemini 2.5. (Reddit)
Kimi K2 Thinking leads agentic coding — tops Tau bench and HLE tool tasks; considered the open-source model closest to GPT-5. (announcement)
DeepSeek Cogito v2.1 — based on DeepSeek-7B, outperforms Mistral-7B and Gemma-9B on introspection tasks, detecting and reporting injected concepts. (announcement)

Issues and Troubleshooting

Manus users can't access TiDB Cloud — quota exhaustion and no console access; ticloud CLI attempts fail without API keys. (TiDB Cloud)
GPT-5.1 connection errors in Aider — response.reasoning.effort validation errors; adjusting reasoning-effort (low/medium/high) doesn't resolve it, possibly related to OpenAI API changes. (LiteLLM issue)
Manus credit system changes — shift to monthly resets confuses users with prior "never expires" plans. (Manus)
Windsurf fixes Gemini 3 loading glitch — update to latest version. (Discord)
Cursor Gemini 3 high-demand errors — team suggests Sonnet 4.5 as fallback. (Discord)

Education and Resources

Community member offers an AI coding education course covering core concepts, advanced models, practical applications, and ethics (contact via DM).
HuggingFace smol-course and agents-course certification questions; newcomers welcomed. (courses)
Tinygrad benchmark resources for comparing CPU performance with Torch in test/external. (GitHub)
AI security red-team training resources including jailbreak prompts and AzureAI chat widget testing. (resources)
llama.cpp recommended as a free, open-source local alternative to Ollama. (GitHub)

---

*Source: Easy AI education project*