How-Tos & Practical Guides

An engineering lead at a mid-sized startup saw his team’s coding agent bill grow 6x in two quarters. Not because the work got 6x harder — because nobody was watching. According to LangChain’s Amy Ru in a July 2, 2026 breakdown, this pattern is everywhere right now: teams celebrating token spend as a proxy for productivity, while the actual spend spirals. This guide covers the four-stage framework Ru outlines for bringing that spend under control using LangSmith. ...

If your AI agent isn’t performing the way you want, the instinct is to tweak the prompt, swap the model, or add another tool. Those changes might help — but they’re guesses. According to Vivek Trivedy (Applied Research Lead at LangChain Labs), the systematic approach to agent improvement is fundamentally a data mining problem. And the data you need is already sitting in your traces. This guide walks through the practical framework Trivedy laid out at AI Engineer World’s Fair 2026 and expanded in a July 7 LangChain blog post. ...

OpenAI released GPT-Realtime-2.1 and GPT-Realtime-2.1-mini on July 6, 2026, via the Realtime API. This isn’t a minor version bump — the update delivers meaningful improvements for production voice agent developers: p95 latency slashed by at least 25%, better alphanumeric recognition, improved noise and silence handling, stronger tool use, and a configurable reasoning effort setting. The mini variant provides a lighter, cost-optimized option for high-volume or latency-sensitive workloads. If you’re running voice agents on GPT-Realtime-2 (released May 2026), this guide covers what changed, how to migrate, and how to decide between 2.1 and 2.1-mini. ...

Researchers at Seoul National University’s CompSec lab published DualView this week (arXiv:2607.03821v1), a deterministic defense against indirect prompt injection (IPI) attacks in personal and agentic AI systems. The paper specifically implements DualView as an OpenClaw plugin — making it directly deployable for OpenClaw users who want to close one of the most difficult security gaps in agentic AI today. The results are striking: 100% IPI block rate on benchmark attacks, with only a ~6.4 percentage point utility loss. No changes to core agent logic required. ...

Claude Code v2.1.202 shipped on July 6, 2026, and the release notes have been generating significant coverage — including some that conflates features from this release with the prior v2.1.198 (July 1). Let’s be precise about what actually landed in each version, because both are genuinely interesting. One important correction to circulating coverage: The /dataviz skill — the chart and dashboard design guidance tool with a runnable color-palette validator — debuted in v2.1.198 on July 1, not in v2.1.202. If you haven’t upgraded recently, you may have just gotten both at once, but they’re separate releases. ...

On July 6, 2026, Illinois Governor JB Pritzker signed SB 315 into law, making Illinois the first US state to require frontier AI developers to undergo annual independent safety audits. The legislation — now officially Public Act 104-0538, the Artificial Intelligence Safety Measures Act — passed with overwhelming bipartisan support: 52-5 in the Senate and 110-0 in the House. It takes effect January 1, 2027. If you’re building or deploying frontier AI systems and you do business in Illinois, here’s what you need to know. ...

If you’ve ever shipped an AI agent to production and thought “I hope it doesn’t accidentally delete something” — this tool is for you. MakerChecker is a new open-source security layer for AI agents that combines a static code scanner with a runtime governance engine. It showed up as a “Show HN” post and immediately generated meaningful discussion. The concept is straightforward: most agent security problems stem from tools that can do dangerous things (move money, delete data, run shell commands) without any human checkpoint. MakerChecker puts checkpoints back in. ...

If you’re running a production MCP server with sticky sessions, a shared session store, or Mcp-Session-Id header handling — pay attention. The MCP 2026-07-28 release candidate is the largest revision to the protocol since its launch, and the headline change is a breaking one: sessions are gone entirely. The final specification ships July 28, 2026. If you haven’t started preparing your MCP server implementations, you have less than a month. ...

Tencent just made a significant move in the open-weights AI race. The full Hunyuan Hy3 model is officially live — and after months of preview builds, the final release is everything the preview promised and more. What Is Hy3? Hy3 is a 295B total parameter Mixture-of-Experts (MoE) model from the Tencent Hy Team. MoE means not all those parameters are active at once: only 21B parameters activate per token, plus a 3.8B MTP (Multi-Token Prediction) layer. The result is a model with frontier-class capability but substantially lower compute costs than dense 295B models. ...

Claude Code v2.1.200 shipped a breaking change that caught a lot of automated workflows by surprise: the default permission mode is now Manual. If you had scripts, CI jobs, or headless agent pipelines relying on the old default behavior — where Claude would proceed without explicit permission approval — those workflows may now be stalling waiting for input that never comes. This guide covers exactly what changed, what it means for your setup, and how to update your configuration so automated workflows keep running. ...