LLMs

My notes from the Open-Source LLM Builders Summit in Lausanne.

In addition to using AI in Obsidian, I'm also allergic to voice-to-text.

MCP Tool Poisoning on the main website for The OWASP Foundation. OWASP is a nonprofit foundation that works to improve the security of software.

I stopped grading three answers myself.

Build and deploy an agent on AWS with Strands and AgentCore

Researchers [Charles Ye], [Jasmine Cui], and [Dylan Hadfield-Menell] have shown that AI Large Language Models (LLMs) can fail to correctly distinguish between different instruction sources because …

Abacus.AI and the case for unified AI workflows

Improve coding agent productiveness with refactored code

From a raw idea to merged code — multi-model LLM councils plan, Ralph loops perfect, and OpenCode worktrees ship every change. Free, open source, runs locally.

When it comes to deploying AI in their projects, many software development teams are flying blind.

When an agent does something, the whole company should learn from it, so that every developer gets access to the shared insights and successes.

The following article originally appeared on Angie Jones’s LinkedIn page and is being republished here with the author’s permission.I'm fascinated by the

The following article originally appeared on Addy Osmani’s blog and is being reposted here with the author’s permission.Loop engineering is replacing

Using DSPy to automatically create, evaluate, and optimize your prompts

Learn how GitHub built Qubot, our internal Copilot-powered analytics agent, to allow any GitHub employee to ask questions about our data in plain language.

A terminal-based coding agent

Perplexity's search as code makes professional AI retrieval mainstream. Discover how a simple implementation yields a massive impact.

AI is flooding teams with pull requests that are too large for humans to review. Qodo's new tools learn your code standards and catch cross-repo bugs before merge.

A new framework, Arbor, they claim, preserves hypotheses, experiments, and lessons learned across long-running research tasks, delivering 2.5x better performance than other models under the same budget.

By Winston Chou, Adrien Alexandre, Lars Olds, Yi Zhang, Garrett Hagemann, and Nathan Kallus

Like Next.js for web apps, but for agents. Markdown for instructions and skills, TypeScript for tools. Durable by default.

How I bridged my MacBook Air and Framework 13 using LM Studio's LM Link to run powerful local models, avoid token costs, and gain portability.

We break down the technical architecture behind our multi-stage vulnerability discovery harness and automated triage loop. Learn how we manage state controls, squash false positives through adversarial review, and route around LLM context limits.

Earlier this year i was studying some open source code. From projects like: Wild, Linux, binutils, MetaCall. I used this `ASIDE.md to help me, giving the LLM the oppurtiunity to teach me while i *have* to write the answers myself.

The primitive for AI agents and the systems they operate. Stateful actors, long-running workflows, and an operating system for agents — run on Rivet Cloud or inside your own VPC.

Perform efficient data retrieval of personal knowledge

A context layer turns warehouse schema, metrics, and business docs into one governed place — so AI data agents like Claude or Codex query your stack reliably.

Stop re-computing the same context. Learn how to build a C++ runtime with copy-on-fork KV snapshots to eliminate redundant LLM prefills in multi-agent pipelines.

Six layers between your LLM and a production agent

Usage-based billing didn’t make agents expensive. It made their existing costs visible, and visibility turns agent economics into a governance problem.