LLM
33 items
26 posts, 7 tools
Agent memory is five layers: history, compaction, semantic recall, working memory and files. How Mastra, LangGraph, OpenAI and Claude implement each.
ACE vs ALTK-Evolve for agent memory: both turn trajectories into lessons, but calibrated delivery matches accuracy on AppWorld at a fraction of the tokens.
Fine-tuned coding agents collapse when you switch scaffolds, from 52.6% to 8.4% on SWE-bench Verified. DCAS shows how to train planning into the model.
OpenJDK's interim AI policy bans code generated in full or in part by LLMs. What it still allows, how it is enforced, and what it means for Java contributors.
SkillSV is a Shapley framework that values the rules, examples and scripts inside an agent skill, cutting skills to 69% of tokens without significant loss.
SIGIL compiles prose SKILL.md files into typed agent harnesses. Prose runs execute 56% of mandated steps; compiled harnesses reach 86% with 0.58x the tokens.
AI API token relays resell OpenAI, Anthropic, and Google access at up to 97.8% off. How the gray market works and what it means for developers on AI APIs.
Debian LLM policy debate: four proposals on AI-generated contributions, from an outright ban to full acceptance, and the fault lines in the HN discussion.
Gemma 4 26B on a 2013 Xeon with no GPU: how a sub-$300 box reached about 5 tokens/sec, the silent MoE bug behind the gibberish, and the upstream fix.
Mesh LLM pools GPUs across home lab machines behind one OpenAI-compatible API, splitting 235B models over iroh and QUIC with no RDMA or NVLink required.
Self-host LiteLLM or Kong, or use a managed gateway like Portkey, OpenRouter, or Cloudflare AI Gateway? A factual breakdown of cost, control, and ops tradeoffs.
Program-as-Weights compiles prompts into compact local adapters that act as fuzzy functions. What the paper claims and where developers could use it first.
Open-source LLM engineering platform: tracing, evals, prompt management, and datasets. Self-hostable, OpenTelemetry-native, with 50+ framework integrations.
Prompt injection through a bank transfer memo: how a €0.02 transfer compromised a banking AI assistant, and the five defense layers agent builders need.
Mastra review and setup for TypeScript AI agents - verified create commands, Mastra Studio, workflows and memory, plus pricing and who should skip it.
OpenRouter review and setup - one API key for 300+ models with fallbacks and provider routing, what it costs, and when to go direct to a provider instead.
LiteLLM vs Portkey vs OpenRouter compared for production AI apps: cost tracking, fallbacks, caching, and load balancing, plus when to use each router.
How KV caching speeds up LLM inference - the math, the code, the memory tradeoffs, and when it stops helping. Every dev running local models hits this wall.
Mercury 2 developer guide: the OpenAI-compatible API, tool use, structured outputs, reasoning levels, and when a diffusion LLM beats an autoregressive one.

Get Smarter About AI Dev
New tutorials, open-source projects, and deep dives on coding agents - delivered weekly.