Technology On the Net

Most production LLM apps optimize for answering. The model always returns something, the UI always shows it, and the user is left to decide whether it is trustworthy. That default fails in the exact places where wrong answers hurt: medical triage assistants that invent dosages, support bots that invent policy exceptions, analytics agents that invent SQL columns, and retrieval apps that answer when the corpus has nothing relevant. Image: Pearson Scott Foresman via Wikimedia Commons (Public domain) Abstention is the deliberate choice to not answer, or to answer only after a human looks. A confidence gate is the machinery that decides when abstention is safer than a guess. This post covers how to build those gates in a production app: what signals to use, where to put the check in the request path, how to phrase a useful decline, and how to measure whether the gate is helping or just annoying users. This is not the same as content policy refusals. Policy refusals block disal...

Ambiguous user messages are one of the quietest ways production LLM assistants waste money and trust. The model fills in a missing detail, the user gets an answer to the wrong question, and the next turn is a correction that burns another full completion. In chat logs this looks like "normal" multi-turn conversation. In practice it is a failure mode you can detect and interrupt. Image: Oregon Department of Transportation via Wikimedia Commons (CC BY 2.0) Image: Lapollard via Wikimedia Commons (CC BY-SA 4.0) A clarification turn is a short, deliberate pause: the assistant asks one targeted question, waits for the answer, then continues with a grounded request. Done well, it cuts wrong-path tool calls, reduces regenerate clicks, and makes the product feel careful instead of overconfident. This guide covers when to ask, how to ask only once, how to keep the rest of the stack from racing ahead, and how to measure whether clarification is helping. Why guess...

Long agent runs do not belong on a single open HTTP request. When a tool loop can take minutes—scraping sites, calling three vendors, waiting on a human approval—you need an async job pattern: accept the work, return a job id immediately, then deliver the result with polling or a webhook. This guide covers a practical design you can ship in October 2026 without inventing exotic infrastructure. Image: mikemacmarketing / photo on flickr via Wikimedia Commons (CC BY 2.0) Image: Midjourney; prompt suggested by Grok via Wikimedia Commons (Public domain) Why synchronous LLM APIs break under real agent work A chat completion that returns in two seconds fits a request/response model. An agent that plans, calls tools, retries, and summarizes does not. Holding the client connection open creates three failure modes: Proxy and load-balancer timeouts cut the stream mid-run even though your worker is still healthy. Client disconnects leave orphaned GPU and tool work unless...

LLM agents that call tools look great in demos and fall apart in production for boring reasons: a payment API times out after the charge succeeded, a search tool returns malformed JSON, or two parallel tool calls race and double-book a resource. The model is rarely the whole problem. The glue around tool calling is. Image: mikemacmarketing / photo on flickr via Wikimedia Commons (CC BY 2.0) Image: Midjourney; prompt suggested by Grok via Wikimedia Commons (Public domain) This post is a practical checklist for making tool use reliable in production agents: how to design tools for retries, when to use idempotency keys, how to validate schemas before the model sees a result, and how to structure the conversation so a failed call does not poison the next turn. What breaks in production (and what does not) When an agent fails a tool call, teams often blame the model. In practice, the failure modes cluster into four buckets: Transport and timeouts. The HTTP clien...

Google ADK for Kotlin 1.0: A Practical Guide to Production AI Agents Google Agent Development Kit (ADK) for Kotlin 1.0 reaches general availability in September 2026, bringing Google’s agent-building framework to Kotlin Multiplatform and Android-first development. The release is aimed at developers who need more than a chat interface: multi-agent workflows, tool execution, approval steps, resumable sessions, local inference, cloud models, and deployment through Vertex AI Agent Platform. This guide explains how to choose ADK for Kotlin, structure an agent application, persist sessions, handle human confirmation safely, and decide when Python or JavaScript ADK remains the better option. Image: mikemacmarketing / photo on flickr via Wikimedia Commons (CC BY 2.0) Image: Midjourney; prompt suggested by Grok via Wikimedia Commons (Public domain) What ADK for Kotlin is designed to solve An AI agent is an application that can interpret a request, decide which action...

Native Multimodal Models for AI Agents: What Ant Group Ling-3.0-flash-VL Means for Visual Feedback Loops AI agents are moving beyond text interfaces. In 2026, many useful agents will need to inspect screenshots, compare visual states, read documents, understand charts, operate software, and verify whether an action produced the intended result. Ant Group’s Ling-3.0-flash-VL is an example of the open multimodal models teams are evaluating for this role. Its reported 124-billion-parameter mixture-of-experts design and 256K context window point toward a model built for long, visually grounded workflows rather than isolated image questions. Image: mikemacmarketing / photo on flickr via Wikimedia Commons (CC BY 2.0) Image: Midjourney; prompt suggested by Grok via Wikimedia Commons (Public domain) The important question is not simply whether a model can describe an image. Agent systems need a more demanding capability: they must observe a changing environment, compar...