3 min read
Indirect Prompt Injection: Tool Output Is Not Instructions
A retrieved document, an email, or a tool result can tell the model to take an action. Delimiters do not stop it. The tool allowlist after untrusted text does.
Tag
LLM
Articles tagged LLM on The Augmented Dev.
9 articles
5 min read
Copilot Auto Model Selection Now Has Tiers: Efficiency, Balance, Intelligence
GitHub Copilot's Auto picker gained efficiency, balance, and intelligence tiers in September 2026, alongside GPT-6 Astra and GPT-6.1 Sol. How the tiers change the cost-quality trade, and what admins control.
7 min read
GitHub Copilot AI Model Comparison: Which Model to Use for Each Task
AI model comparison for GitHub Copilot Chat: GPT-5.6, Claude, Gemini, Grok, and Kimi, plus when Auto is the cheaper default.
3 min read
Put the Workflow in Code When the Agent Keeps Improvising
Which steps belong in a fixed program and which belong in a model, so an agent stops skipping the check that production depends on.
7 min read
Evaluating LLM Applications: Getting Past 'It Looks Good to Me'
Shipping LLM features on vibes works until a prompt tweak silently breaks ten other cases. Here is how to build evals that catch regressions before your users do.
7 min read
Getting Structured Output From an LLM Without the Heartbreak
Asking a model to 'return JSON' and parsing the result is how you get 3 a.m. pages. Tool schemas, constrained decoding, and validation turn a probabilistic text generator into a reliable API.
7 min read
Chunking Strategies for RAG: Where Retrieval Quality Is Won or Lost
Most RAG systems that retrieve bad context aren't failing at embeddings or reranking — they're failing at chunking. How you split documents quietly decides what your model can ever find.
8 min read
Designing a Production-Ready RAG System: What Actually Matters
Retrieval-augmented generation demos look great. Production RAG is a different engineering problem — chunking, hybrid retrieval, reranking, evaluation, and the failure modes nobody mentions at conferences.
8 min read
Event-Driven Architecture for LLM Pipelines: Queues, Retries, and the Cost of Synchronous Chains
LLM-powered pipelines break in the same ways old monoliths did — when you stack synchronous calls. Queues, events, and the patterns that make a six-step AI pipeline survive a partial outage.