Notes from the foundry
Engineering essays on generative AI, retrieval systems, and what it takes to ship intelligent software to production.
Decoding Biology: How Kimi K3 is Accelerating Medical Diagnostics and Genomic Research
Discover how clinical researchers are using Moonshot AI's Kimi K3 model to analyze entire patient medical histories alongside thousands of pages of genomic research to diagnose rare diseases.
From Video to Insights: Deep Reasoning with Kimi K3's Native Multimodal Engine
Learn how Kimi K3's native multimodal capabilities allow marketers and researchers to extract profound insights directly from hours of raw video and audio.
Revolutionizing Quantitative Finance: Kimi K3 as the Ultimate Algorithmic Trading Architect
An in-depth exploration of how Moonshot AI's Kimi K3 is transforming quantitative finance. Learn how hedge funds use its 1-million token context to analyze years of market data and generate complex trading algorithms.
Claude Fable 5 vs Kimi K3: Choosing the Right Frontier AI for Enterprise Workflows
A detailed comparison of Anthropic's Claude Fable 5 and Moonshot AI's Kimi K3. We analyze safety classifiers, 1-million token context capabilities, coding, and API pricing.
Kimi K3 vs GPT-5.6 Sol: The Battle for Long-Context Reasoning in 2026
A comprehensive, deep-dive comparison between Moonshot AI's Kimi K3 and OpenAI's GPT-5.6 Sol. We analyze architecture, 1-million token context performance, coding capabilities, and enterprise pricing.
AI Code Execution on Cloudflare Sandboxes
Run untrusted agent-generated code next to your app on Cloudflare, with no separate container host or VPC. Building the tool, end to end.
Hardening an AI Code Sandbox: Seccomp, Egress Firewalls, and gVisor
A locked-down container stops the obvious attacks. Here are the controls that stop a determined one: a non-root user, a tight seccomp profile, resource ulimits, a network egress allowlist, and a real kernel boundary with gVisor.
Give a Coding Agent a Persistent Sandbox: Clone, Edit, Test, PR
One-shot code execution can't build software. An autonomous coding agent needs a filesystem that survives across turns, to install dependencies, edit files, run the test suite, read the failures, and open a pull request. This builds that with a persistent E2B sandbox.
Building a Realtime Voice Support Agent with Vercel AI SDK v7
Voice agents used to mean marrying one provider's WebSocket event format. I used the new experimental realtime API in Vercel AI SDK v7 to build an order-status voice agent I can move between providers.
Lightweight AI Sandboxes with WebAssembly: Pyodide and QuickJS
When you can't run a container, serverless functions, the edge, cold-start-sensitive paths, you can still isolate untrusted code in-process with WebAssembly. This runs LLM-written Python via Pyodide and JavaScript via QuickJS, each with hard memory and time limits.
Why AI Agents Need Sandboxes (and How to Build One with Docker)
The moment you let an LLM run the code it writes, you've handed a stranger a shell on your server. Here's the threat model, and a step-by-step build of an isolated Docker sandbox your agent can safely execute code in.
Building an Autonomous Lead Generation Agent using Eve.dev
Discover how to build an AI sales assistant that scores leads, crafts personalized emails, and uses durable wait states to manage multi-day follow-up sequences using Eve.dev.
How to Build a Durable AI Support Agent with Eve.dev and Next.js
Learn how to use Eve.dev to build a durable, crash-resistant AI customer support agent that can pause, wait for user input, and resume seamlessly within a Next.js application.
Creating a File-System Based Agentic Workflow with Eve.dev
Discover how Eve.dev's unique filesystem-first architecture allows you to organize complex agent capabilities, skills, and prompts just like a Next.js application.
Automate Code Reviews with a Durable Eve.dev GitHub Agent
Learn how to build an autonomous AI code reviewer using Eve.dev that listens to GitHub webhooks, analyzes pull requests, and posts detailed inline comments.
Automating Employee Onboarding with an Eve.dev HR Agent
Learn how to use Eve.dev to build an intelligent Human Resources assistant that guides new hires through onboarding, collects documents, and durably waits for responses over weeks.
Implementing Human-in-the-Loop AI Agents using Eve.dev and TypeScript
Ensure safety and accuracy in your AI workflows. Learn how to build human-in-the-loop authorization gates using Eve.dev's durable execution framework.
Building a Resilient Web Scraping Agent with Eve.dev and Puppeteer
Learn how to build a durable AI web scraping agent that extracts structured data, survives browser crashes, and handles rate limiting automatically using Eve.dev.
Create an AI Social Media Manager Agent with Eve.dev
Build an autonomous AI social media manager that researches trends, generates week-long content schedules, and durably posts content over time using Eve.dev.
Grok Voice Agent Builder Review: Can a No-Code Tool Ship a Production Voice Agent?
A hands-on review of xAI's Grok Voice Agent Builder beta. What the no-code voice agent platform handles well, where you still need the API, and how to decide which one your project needs.