Article archive
Every article, newest first.
- Haiku 5.5: Cheaper AI can lose the evidence your strongest model needs 2026-10-11
- Mistral Large 4: Smarter reasoning cannot fix a misread document 2026-10-11
- Reflection’s Beam: Shorter reasoning only saves money if the answer still works 2026-10-11
- EmbeddingGemma 2: Better AI answers can start with finding the missing evidence 2026-10-11
- DeepSeek V4.1 Flash: Cheaper AI memory could change what your model remembers 2026-10-11
- GPT-6.1 Sol: The biggest AI saving may come from changing your default model 2026-10-11
- Gemini Robotics 2: A new robot should not mean starting AI training again 2026-10-11
- Physical Intelligence’s π0.7: Teach a robot a new task before retraining it 2026-10-11
- Mecka: Robot-training data is worth more when it teaches a skill, not just records one 2026-10-11
- TwelveLabs: The value of AI video labels is the expert work they save 2026-10-11
- Snorkel AI: An agent learns what the reward accepts—not necessarily what the business needs 2026-10-11
- Datology: Better training data could make a smaller model the better buy 2026-10-11
- NVIDIA: Synthetic data needs a quality specification before it needs scale 2026-10-11
- Mistral: A stronger model needs harder lessons—not just more training 2026-10-11
- NVIDIA: Cheaper training arithmetic could buy more experiments—not just bigger models 2026-10-11
- FlashAttention-4: Faster GPUs need smarter software to speed up AI training 2026-10-11
- Thinking Machines and UIUC: Better tests can teach AI what more examples cannot 2026-10-11
- vLLM: Keeping AI answers on time can matter more than generating more tokens 2026-10-11
- SGLang: Smaller AI memory can serve more users—if the extra errors are affordable 2026-10-11
- Inco AI’s DFlash 2 speeds up answers by making better guesses for the same model 2026-10-11
- PrismML’s smaller model could bring useful AI onto hardware you already own 2026-10-11
- OpenAI and Ironclad: One missed business rule can erase an AI agent’s progress 2026-10-11
- Surge AI: A model can get the spreadsheet math right and still give the wrong answer 2026-10-11
- Mercor: Beating accountants on a test is not yet proof of cheaper accounting 2026-10-11
- Harvey’s AI test: One invented fact can cancel an otherwise complete answer 2026-10-11
- Factory: Your old test cases may be the key to trusting AI-written repairs 2026-10-11