YLEM tracks the frontier of machine intelligence as a set of living field guides, a milestone timeline and hands-on lab reports. What changed, why it matters, and what you can build with it now.https://ylem.si/https://ylem.si/lab/minimal-mcp-server/https://ylem.si/lab/minimal-mcp-server/Adopt. We built a 31-line MCP server that searches local Markdown notes and benchmarked it. The protocol overhead is negligible; the real work is tool design.Mon, 05 Oct 2026 00:00:00 GMThttps://ylem.si/log/2026-10-05-registertool/https://ylem.si/log/2026-10-05-registertool/MCP TypeScript SDK: server.tool() is deprecated, use registerTool()Mon, 05 Oct 2026 00:00:00 GMThttps://ylem.si/fields/models/https://ylem.si/fields/models/First draft. Capability gains now come as much from post-training and inference-time compute as from raw pre-training scale.Mon, 05 Oct 2026 00:00:00 GMThttps://ylem.si/fields/reasoning/https://ylem.si/fields/reasoning/First draft. Spending more compute at inference time, through long internal reasoning trained with reinforcement learning, is the most important capability lever since scaling pre-training.Mon, 05 Oct 2026 00:00:00 GMThttps://ylem.si/fields/agents/https://ylem.si/fields/agents/First draft. Agents are reliable inside well-instrumented environments such as codebases and browsers with clear goals. Open-ended autonomy over long horizons is still fragile.Mon, 05 Oct 2026 00:00:00 GMThttps://ylem.si/fields/coding/https://ylem.si/fields/coding/First draft. Coding is the first domain where agents do substantial delegated work. The bottleneck has moved from writing code to reviewing and verifying it.Mon, 05 Oct 2026 00:00:00 GMThttps://ylem.si/fields/multimodal/https://ylem.si/fields/multimodal/First draft. Understanding images and documents is routine, and real-time voice works. Video generation is impressive but hard to control precisely.Mon, 05 Oct 2026 00:00:00 GMThttps://ylem.si/fields/memory/https://ylem.si/fields/memory/First draft. Retrieval-augmented generation is the standard enterprise pattern. The hard problem has shifted from finding text to deciding what an agent should remember.Mon, 05 Oct 2026 00:00:00 GMThttps://ylem.si/fields/robotics/https://ylem.si/fields/robotics/First draft. Vision-language-action models show real generalisation in research. Deployment is held back by data, hardware cost and reliability.Mon, 05 Oct 2026 00:00:00 GMThttps://ylem.si/fields/infrastructure/https://ylem.si/fields/infrastructure/First draft. Inference, not training, increasingly dominates compute demand, driven by reasoning models and agents that use many tokens per task.Mon, 05 Oct 2026 00:00:00 GMThttps://ylem.si/fields/alignment/https://ylem.si/fields/alignment/First draft. Alignment is now part of the product. Techniques like RLHF and constitutional training shape every assistant, while interpretability and evaluation of dangerous capabilities are active frontiers.Mon, 05 Oct 2026 00:00:00 GMThttps://ylem.si/log/2026-10-04-reading-list/https://ylem.si/log/2026-10-04-reading-list/Eight papers on the timeline worth reading in fullSun, 04 Oct 2026 00:00:00 GMThttps://ylem.si/log/2026-10-03-why-fields/https://ylem.si/log/2026-10-03-why-fields/Why YLEM is organised by field, not by dateSat, 03 Oct 2026 00:00:00 GMT