Most AI model releases focus on benchmarks. This one is more interesting because of what Microsoft says it did not use. With MAI-Thinking-1, Microsoft claims no third-party LLM-generated synthetic data during pre-training, active filtering of AI-generated content, and no hidden
@whats_ai
-
Microsoft’s MAI-Thinking-1 uses cleaner data, no teacher model
By
–
New video is live.
Most AI labs use synthetic data or distillation because it is cheap, fast, and effective. Microsoft just made a very different bet with MAI-Thinking-1: cleaner pre-training data, no third-party teacher model, and a much more transparent report than usual. It -
Codex is my go-to backup for these moments
By
–
haha Codex has honestly become my go-to backup for exactly these moments.
-
Thousands of agents in loops overnight is not everyone’s reality yet
By
–
"this is just how engineering is done now" is doing a lot of work haha, thousands of agents in loops overnight is his reality, not exactly everyone's yet.
-
Happy with Codex over Claude Code until Fable returns
By
–
haha honestly been surprisingly happy with Codex over Claude Code lately, at least until Fable's back.
-
Supprimer 80% des prompts améliore l’agent
By
–
"Deleted 80% of our prompt and the agent got better" matches what I keep seeing, every rule you bolt on to patch one case ends up costing you on ten others. I periodically run cleaning runs to remove duplicates and make each skill more concise.
-
Plain RAG with good chunking outperforms graph and agentic RAG
By
–
Honestly plain RAG with good chunking still wins most builds, graph and agentic RAG only earn their keep once retrieval is genuinely your bottleneck.
-
Rent the model, own the system: integration and observability compound
By
–
"Rent the model, own the system" is the cleanest version of this I've seen, the integration and eval and observability layer is the part that actually compounds.
-
Per-task routing becomes default, single model apps feel dated
By
–
Per-task routing is becoming the default, locking your whole app to one model already feels dated.
-
Self-Prompting Systems: Beyond the Prompt Itself
By
–
"Build a system that prompts itself" really lands, writing the prompt was never the hard part, it's everything you wrap around it.