---
title: 'Pydantic Blog: Updates, Observability & AI Insights'
description: >-
  Check out our articles for Pydantic product and feature updates & engineering
  deep dives on topics related to AI, LLMs, observability, agent frameworks &
  more
canonical: 'https://pydantic.dev/articles'
---

> Markdown version of [Pydantic articles](https://pydantic.dev/articles) — the canonical HTML page.
>
> Site index: [/llms.txt](https://pydantic.dev/llms.txt) · RSS: [/feed.xml](https://pydantic.dev/feed.xml)

---

# Pydantic articles

Check out our articles for Pydantic product and feature updates & engineering deep dives on topics related to AI, LLMs, observability, agent frameworks & more.

- [A research agent, three ways](https://pydantic.dev/articles/harness-exa) — 2026-07-21 ([markdown](https://pydantic.dev/articles/harness-exa.md))
  Harness Week, day two: start with Pydantic AI's native WebSearch for the lookup, level up to the ExaAgent capability when the question earns a full multi-step research pass, then compose ExaSearch with the harness pillars when you want the deep research loop to be yours. Exa is the retrieval that powers Cursor and Cognition's Devin.
- [You've built this agent before](https://pydantic.dev/articles/harness-week) — 2026-07-20 ([markdown](https://pydantic.dev/articles/harness-week.md))
  Harness Week: Pydantic AI Harness is the standard library for agents. File access, memory, guardrails, and sub-agents as swappable capabilities, proven by Vstorm's six endorsed packages. Five days, one agent built from parts, with Modal, LocalStack, Macroscope, and Exa.
- [Your traces already know how to fix your prompt](https://pydantic.dev/articles/logfire-prompt-optimization) — 2026-07-17 ([markdown](https://pydantic.dev/articles/logfire-prompt-optimization.md))
  The last mile in Pydantic Logfire: the optimizer reads your production traces and proposes one evidence-cited prompt edit you can copy into any agent, on any framework. And if it's a managed prompt, accepting it is shipping it. No benchmark, no black box, no deploy.
- [Score Freely with Pydantic Logfire](https://pydantic.dev/articles/logfire-annotations) — 2026-07-16 ([markdown](https://pydantic.dev/articles/logfire-annotations.md))
  Annotations in Pydantic Logfire: human reviewers grade agent runs with a three-tier verdict, a failure category, and the corrected answer, in three keystrokes. Structured judgment that outlives the trace and feeds your evals.
- [Logfire joins Stripe Projects](https://pydantic.dev/articles/logfire-stripe-projects) — 2026-07-15 ([markdown](https://pydantic.dev/articles/logfire-stripe-projects.md))
  Pydantic Logfire is now a provider in Stripe Projects. Two CLI commands provision a Logfire project and write working credentials to your environment. No dashboard, no copy-pasting, and your coding agent can do it too.
- [One key in, no keys out](https://pydantic.dev/articles/logfire-ai-gateway) — 2026-07-15 ([markdown](https://pydantic.dev/articles/logfire-ai-gateway.md))
  The Pydantic Logfire AI gateway: one key for every provider going in, and data loss prevention that stops secrets and PII leaking out. Plus failover, load balancing, and hard spending caps, on the same trace as the rest of your agent.
- [Logfire sandboxes - the Full Monty](https://pydantic.dev/articles/logfire-sandboxes) — 2026-07-15 ([markdown](https://pydantic.dev/articles/logfire-sandboxes.md))
  We're building a unified interface to sandbox providers for running Python code, with full REPL support, host callbacks, and billing through Pydantic Logfire.
- [When agents build agents](https://pydantic.dev/articles/when-agents-build-agents) — 2026-07-14 ([markdown](https://pydantic.dev/articles/when-agents-build-agents.md))
  Part two of the series: a loop is a system of agents that picks its own structure, builds its own tools, and outlives a single run. What that looks like with sub-agents, dynamic workflows, and runtime authoring.
- [The average run is lying to you](https://pydantic.dev/articles/logfire-agents-llms-view) — 2026-07-14 ([markdown](https://pydantic.dev/articles/logfire-agents-llms-view.md))
  Two new views in Pydantic Logfire for the agent era: a per-agent view with the run distributions that surface the runaway, and a per-model LLMs inventory with latency, throughput, and cost priced from open data. Both end in the trace that explains it.
- [You perfected the wrong agent](https://pydantic.dev/articles/agents-week) — 2026-07-13 ([markdown](https://pydantic.dev/articles/agents-week.md))
  Evals tell you an agent is right; only production tells you it's the right agent. Agents Week: ship rough, read the traces, and run them like cattle, not pets. The agent and LLM views, the gateway, annotations, the optimizer, and managed prompts.
- [Observability tools agents want](https://pydantic.dev/articles/observability-tools-agents-want) — 2026-07-02 ([markdown](https://pydantic.dev/articles/observability-tools-agents-want.md))
  A hands-on comparison of the MCPs that let AI agents investigate observability data across Pydantic Logfire, ClickStack, LangSmith, Braintrust, Galileo, Arize Phoenix, and Langfuse.
- [Dispatches from the hallway track](https://pydantic.dev/articles/hallway-track-2026) — 2026-06-30 ([markdown](https://pydantic.dev/articles/hallway-track-2026.md))
  On what people are actually saying between the talks: agents, planning, cynicism, and the particular anxiety of learning new things when you're not sure any of it will stick.
- [Some customers can't use your cloud. Now what?](https://pydantic.dev/articles/some-customers-cant-use-your-cloud) — 2026-06-29 ([markdown](https://pydantic.dev/articles/some-customers-cant-use-your-cloud.md))
  What we learned building the self-hosted Logfire Helm chart for teams that cannot send observability data to a vendor-operated cloud.
- [What makes a good harness](https://pydantic.dev/articles/what-makes-a-good-harness) — 2026-06-26 ([markdown](https://pydantic.dev/articles/what-makes-a-good-harness.md))
  What makes an agent harness good is timing: getting the right text to the model at the right moment. What that looks like with capabilities, code mode, and just-in-time tools.
- [A maturity model for applied generative AI: a framework for senior leaders](https://pydantic.dev/articles/applied-generative-ai-maturity-model) — 2026-06-25 ([markdown](https://pydantic.dev/articles/applied-generative-ai-maturity-model.md))
  A five-level, five-dimension maturity model for applied generative AI that senior engineering leaders can hold up against their org honestly.
- [A production-ready full-stack template for AI agents with Pydantic AI and Logfire](https://pydantic.dev/articles/full-stack-ai-agent-template-pydantic-ai) — 2026-06-24 ([markdown](https://pydantic.dev/articles/full-stack-ai-agent-template-pydantic-ai.md))
  An open-source CLI that generates production-ready FastAPI and Next.js apps with Pydantic AI agents, WebSocket streaming, auth, and Logfire tracing built in.
- [Pydantic AI v2: capable agentic loops](https://pydantic.dev/articles/pydantic-ai-v2) — 2026-06-23 ([markdown](https://pydantic.dev/articles/pydantic-ai-v2.md))
  Pydantic AI v2 is here. One primitive, the capability, now bundles an agent's instructions, tools, hooks, and settings, around a small portable core and the Pydantic AI Harness.
- [Skills on steroids: on-demand capabilities in Pydantic AI](https://pydantic.dev/articles/pydantic-ai-capabilities) — 2026-06-22 ([markdown](https://pydantic.dev/articles/pydantic-ai-capabilities.md))
  Mark a Pydantic AI capability defer_loading=True and the model loads its tools, instructions, and hooks only when it asks for them. Leaner context, lower token cost.
- [When you do not know what you are looking for](https://pydantic.dev/articles/logfire-metrics-explorer) — 2026-06-19 ([markdown](https://pydantic.dev/articles/logfire-metrics-explorer.md))
  A new Metrics explorer in Pydantic Logfire. A three-step wizard from namespace to metric to dimension, with a per-dimension preview grid and View SQL on every card so you can graduate to the full editor without retyping.
- [The OOM that was not your agent](https://pydantic.dev/articles/logfire-hosts-view) — 2026-06-18 ([markdown](https://pydantic.dev/articles/logfire-hosts-view.md))
  A new Hosts view in Pydantic Logfire with CPU, memory, load, disk, and network per host, powered by the standard OpenTelemetry hostmetrics receiver, next to the application traces those hosts are producing.
- [The pod that did not survive Tuesday's deploy](https://pydantic.dev/articles/logfire-kubernetes-view) — 2026-06-17 ([markdown](https://pydantic.dev/articles/logfire-kubernetes-view.md))
  A new Kubernetes view in Pydantic Logfire. Cluster, namespace, workload, pod, node and image inventories powered by the standard OpenTelemetry receivers, alongside the traces from the workloads they describe.
- [When the slow thing is not the model](https://pydantic.dev/articles/logfire-services-view) — 2026-06-16 ([markdown](https://pydantic.dev/articles/logfire-services-view.md))
  A new Services view in Pydantic Logfire that acts as the navigation layer for the live view: RED metrics per service, a topology graph drawn from your traces, deployment markers, one click into the failing trace, even when you operate thousands of services.
- [Agents are the new services](https://pydantic.dev/articles/agents-are-the-new-services) — 2026-06-15 ([markdown](https://pydantic.dev/articles/agents-are-the-new-services.md))
  Production agents fail in shapes classical service observability does not catch. Here is the comprehensive picture we are building in Pydantic Logfire, across every agent framework, and the four new views shipping this week.
- [Simpler metric queries in Logfire](https://pydantic.dev/articles/simpler-metric-queries) — 2026-06-12 ([markdown](https://pydantic.dev/articles/simpler-metric-queries.md))
  New metric_* SQL functions make querying OpenTelemetry metrics in Logfire dramatically simpler: histogram percentiles in one line, with the same query working for every metric type.
- [A $10 starter credit for Pydantic AI Gateway](https://pydantic.dev/articles/ai-gateway-starter-credit) — 2026-06-12 ([markdown](https://pydantic.dev/articles/ai-gateway-starter-credit.md))
  Team and Growth organizations now get a one-time $10 Pydantic AI Gateway starter credit, including when a Personal account is converted into an organization.
- [Cost control, reframed: "AI spend is the new headcount"](https://pydantic.dev/articles/ai-spend-is-the-new-headcount) — 2026-06-09 ([markdown](https://pydantic.dev/articles/ai-spend-is-the-new-headcount.md))
  AI spend scales like headcount, not SaaS: usage-based, salary-sized, and ungoverned. Why cost control is really an observability question.
- [How and why to run modified Python code using the ast module](https://pydantic.dev/articles/eval-type-backport) — 2026-06-08 ([markdown](https://pydantic.dev/articles/eval-type-backport.md))
  Learn how to use Python's ast module to transform and execute modified code, using eval-type-backport as a real-world example.
- [The best AI observability platform in 2026: top picks for building in production](https://pydantic.dev/articles/best-ai-observability-platform) — 2026-06-04 ([markdown](https://pydantic.dev/articles/best-ai-observability-platform.md))
  A practical guide to choosing an AI observability platform in 2026, with honest picks for teams building production AI applications.
- [SELECT * FROM clickbait()](https://pydantic.dev/articles/select-star-from-clickbait) — 2026-06-03 ([markdown](https://pydantic.dev/articles/select-star-from-clickbait.md))
  Find common ground with your agents by talking to them in a language they already speak.
- [The harness thesis](https://pydantic.dev/articles/the-harness-thesis) — 2026-06-03 ([markdown](https://pydantic.dev/articles/the-harness-thesis.md))
  Long-horizon AI agents need more than smarter models. They need a harness that makes agent work observable, governable, and reusable.
- [AI observability is the most underrated leverage for agent products](https://pydantic.dev/articles/atlas-ai-observability-for-agent-products) — 2026-06-02 ([markdown](https://pydantic.dev/articles/atlas-ai-observability-for-agent-products.md))
  How Atlas, eight engineers and no data analyst, turned product questions and bug hunts into one-line prompts using OpenTelemetry baggage, Pydantic Logfire, and two basic agent skills.
- [Welcoming Bill Easton as Head of Product](https://pydantic.dev/articles/bill-easton-head-of-product) — 2026-06-01 ([markdown](https://pydantic.dev/articles/bill-easton-head-of-product.md))
  Bill Easton joins Pydantic as Head of Product, bringing deep experience in observability, developer tools, and AI infrastructure from Elastic, Amazon, and the open-source community.
- [Observe Codex with Logfire](https://pydantic.dev/articles/codex-logfire-plugins) — 2026-06-01 ([markdown](https://pydantic.dev/articles/codex-logfire-plugins.md))
  Export Codex sessions to Logfire, then let Codex query app telemetry through MCP to debug a real PydanticAI regression.
- [Hack Monty 2: a $10,000 bounty to break our Python sandbox](https://pydantic.dev/articles/hack-monty-2) — 2026-05-30 ([markdown](https://pydantic.dev/articles/hack-monty-2.md))
  Hack Monty is back. Together with Prefect and Hugging face, we're putting up a $10,000 bounty for anyone who can break out of the Pydantic Monty sandbox and read the secret file or environment variable. Here's how to play.
- [Emscripten wheels for Pydantic](https://pydantic.dev/articles/emscripten-wheels-pydantic) — 2026-05-29 ([markdown](https://pydantic.dev/articles/emscripten-wheels-pydantic.md))
  A complete guide on how to build and test Emscripten wheels
- [Pydantic Logfire vs LangSmith: AI observability that traces your whole app](https://pydantic.dev/articles/langsmith-alternative-python) — 2026-05-28 ([markdown](https://pydantic.dev/articles/langsmith-alternative-python.md))
  Compare Pydantic Logfire and LangSmith for AI observability. Pricing, features, migration stories, and why teams switch for full-stack tracing across the whole app.
- [SLO monitoring in Logfire](https://pydantic.dev/articles/slo-monitoring-logfire) — 2026-05-27 ([markdown](https://pydantic.dev/articles/slo-monitoring-logfire.md))
  A practical guide to monitoring Service Level Objectives in production. From SLIs and error budgets to multi-window, multi-burn-rate alerts implemented in Pydantic Logfire with SQL.
- [Announcing Logfire Enterprise Dedicated: Single-Tenant Observability, Managed by Pydantic](https://pydantic.dev/articles/logfire-enterprise-dedicated) — 2026-05-26 ([markdown](https://pydantic.dev/articles/logfire-enterprise-dedicated.md))
  Logfire Enterprise Dedicated gives teams a fully managed, single-tenant deployment of Logfire with a dedicated VPC, database, and compute in a region you choose, operated by Pydantic.
- [The runtime layer underneath your Pydantic AI agent](https://pydantic.dev/articles/runtime-layer-pydantic-ai-kitaru) — 2026-05-25 ([markdown](https://pydantic.dev/articles/runtime-layer-pydantic-ai-kitaru.md))
  How Kitaru by ZenML adds durable execution: checkpoints, replay, resume, wait states, and artifacts underneath a Pydantic AI agent, without changing the agent itself.
- [Welcoming Gregor Stewart as a Pydantic Advisor](https://pydantic.dev/articles/gregor_stewart_advisor) — 2026-05-20 ([markdown](https://pydantic.dev/articles/gregor_stewart_advisor.md))
  Gregor Stewart, Chief AI Officer at SentinelOne, joins Pydantic as an advisor — bringing deep experience at the intersection of AI and production systems.
- [Launch AI tools without storing API keys](https://pydantic.dev/articles/logfire-gateway-launch) — 2026-05-18 ([markdown](https://pydantic.dev/articles/logfire-gateway-launch.md))
  Run Claude Code, Codex, and other AI coding tools through the Logfire Gateway with one command. No plaintext API keys on disk, hierarchical spend limits, and cross-provider routing.
- [Teaching DataFusion to read struct fields efficiently](https://pydantic.dev/articles/struct-field-pushdown) — 2026-05-14 ([markdown](https://pydantic.dev/articles/struct-field-pushdown.md))
  How we made struct field access in Parquet as efficient as top-level column access, from filter pushdown to leaf-level projection to row group pruning
- [Build durable agents with Restate and Pydantic AI](https://pydantic.dev/articles/restate-durable-execution-pydanticai) — 2026-05-12 ([markdown](https://pydantic.dev/articles/restate-durable-execution-pydanticai.md))
  Restate now integrates with Pydantic AI, adding durable execution, stateful sessions, human-in-the-loop, and multi-agent orchestration in a few lines of code.
- [FastA2A is moving to Datalayer](https://pydantic.dev/articles/pydantic-donates-fasta2a-to-datalayer) — 2026-05-08 ([markdown](https://pydantic.dev/articles/pydantic-donates-fasta2a-to-datalayer.md))
  Pydantic is donating the FastA2A repository to Datalayer, who will take over maintenance and drive new development of the A2A protocol implementation.
- [Hack Monty - Postmortem](https://pydantic.dev/articles/hack-monty-postmortem) — 2026-05-07 ([markdown](https://pydantic.dev/articles/hack-monty-postmortem.md))
  A postmortem of the Hack Monty security disclosure, which revealed a vulnerability in the Pydantic Monty Python sandbox. We discuss the details of the vulnerability, how it was discovered, and the steps taken to address it.
- [Why we ditched database triggers for audit logs](https://pydantic.dev/articles/audit-logs-replace-database-triggers) — 2026-05-05 ([markdown](https://pydantic.dev/articles/audit-logs-replace-database-triggers.md))
  How we migrated Logfire's audit logging from PostgreSQL triggers to an async, Redis-backed architecture — and the incident that made us do it.
- [Introducing online evals in Pydantic Logfire](https://pydantic.dev/articles/online-evals-pydantic-logfire) — 2026-04-30 ([markdown](https://pydantic.dev/articles/online-evals-pydantic-logfire.md))
  Pydantic Logfire now supports online evals: automated quality monitoring for production AI agents, with configurable judges, built on OpenTelemetry.
- [AI Is Making Remote Work a Bigger Advantage Than Ever](https://pydantic.dev/articles/remote-work-ai-age) — 2026-04-23 ([markdown](https://pydantic.dev/articles/remote-work-ai-age.md))
  Why being a distributed team has become a strategic advantage in 2026, and how AI is resolving the tradeoffs that used to hold remote work back.
- [Hack Monty: a $5,000 bounty to break our Python sandbox](https://pydantic.dev/articles/hack-monty) — 2026-04-20 ([markdown](https://pydantic.dev/articles/hack-monty.md))
  Pydantic Monty is offering a $5,000 bounty to anyone who can break out of the sandbox and read the secret file or environment variable. Here's how to play.
- [Engineering MCP tools for token efficiency](https://pydantic.dev/articles/engineering-mcp-tools-for-token-efficiency) — 2026-04-16 ([markdown](https://pydantic.dev/articles/engineering-mcp-tools-for-token-efficiency.md))
  A practical guide to optimizing MCP server tool definitions for reduced token usage, using snapshot testing in Rust
- [Smarter bloom filters in Parquet: how filter folding speeds up point lookups in Logfire](https://pydantic.dev/articles/bloom-filter-folding-parquet-logfire) — 2026-04-15 ([markdown](https://pydantic.dev/articles/bloom-filter-folding-parquet-logfire.md))
  We contributed bloom filter folding to Apache Arrow, making Parquet bloom filters smaller and more effective for Logfire queries.
- [Vercel AI SDK and Logfire: what happens when everyone speaks OpenTelemetry](https://pydantic.dev/articles/vercel-ai-sdk-logfire-otel) — 2026-04-14 ([markdown](https://pydantic.dev/articles/vercel-ai-sdk-logfire-otel.md))
  Full LLM observability for Vercel AI SDK apps with Pydantic Logfire, powered by OpenTelemetry. Conversation replay, token usage, cost tracking, and streaming latency from a few lines of config.
- [Announcement: Pydantic v2.13 Release](https://pydantic.dev/articles/pydantic-v2-13-release) — 2026-04-13 ([markdown](https://pydantic.dev/articles/pydantic-v2-13-release.md))
  Pydantic v2.13 introduces polymorphic serialization, exclude_if for computed fields, ascii_only validation, and private attribute factories.
- [Full-stack Kubernetes observability with Logfire](https://pydantic.dev/articles/kubernetes-cluster-observability-logfire) — 2026-04-10 ([markdown](https://pydantic.dev/articles/kubernetes-cluster-observability-logfire.md))
  How to collect Kubernetes cluster metrics, enrich application traces with pod metadata, and build unified dashboards in Logfire using the OpenTelemetry Collector.
- [Your agent would rather write code](https://pydantic.dev/articles/your-agent-would-rather-write-code) — 2026-04-08 ([markdown](https://pydantic.dev/articles/your-agent-would-rather-write-code.md))
  We built 40+ MCP tools for Logfire. Then we let the agent write Python instead, and it was immediately better. Here's what happened.
- [How curvy is your data: an investigation into Hilbert curve sorting](https://pydantic.dev/articles/how-curvy-is-your-data) — 2026-04-07 ([markdown](https://pydantic.dev/articles/how-curvy-is-your-data.md))
  We tried Hilbert curve sorting to speed up Logfire queries. It regressed performance. Here's what we learned about data layout and row group pruning
- [Lies, Damn Lies, and the Box Model (Part 3)](https://pydantic.dev/articles/lies-damn-lies-and-the-box-model-part-3) — 2026-04-03 ([markdown](https://pydantic.dev/articles/lies-damn-lies-and-the-box-model-part-3.md))
  On design process, comforting fictions, and imagining what software could become.
- [Lies, Damn Lies, and the Box Model (Part 2)](https://pydantic.dev/articles/lies-damn-lies-and-the-box-model-part-2) — 2026-04-02 ([markdown](https://pydantic.dev/articles/lies-damn-lies-and-the-box-model-part-2.md))
  We rebuilt Logfire's navigation from scratch. Here's what the process actually looked like.
- [Lies, Damn Lies, and the Box Model (Part 1)](https://pydantic.dev/articles/lies-damn-lies-and-the-box-model-part-1) — 2026-04-01 ([markdown](https://pydantic.dev/articles/lies-damn-lies-and-the-box-model-part-1.md))
  On debugging a phantom scrollbar with an AI, and why CSS box model bugs still humble us in 2026.
- [The Unreasonable Cost Effectiveness of Pydantic Logfire](https://pydantic.dev/articles/ai-observability-pricing-comparison) — 2026-03-31 ([markdown](https://pydantic.dev/articles/ai-observability-pricing-comparison.md))
  Compare AI observability pricing across Logfire, LangSmith, Langfuse, and Arize AX. See why the billing unit matters more than the headline price — with real cost breakdowns at 5M, 50M, and 500M spans.
- [Building IaC providers for Logfire: design decisions that mattered](https://pydantic.dev/articles/infrastructure-as-code-provider-for-logfire) — 2026-03-30 ([markdown](https://pydantic.dev/articles/infrastructure-as-code-provider-for-logfire.md))
  Observability as Code workflows with Terraform, OpenTofu, and Pulumi providers for Pydantic Logfire. What we learned about modeling observability resources as code.
- [From Prompts to Processes: Full-Stack Agent Visibility with AgentSH + Pydantic Logfire](https://pydantic.dev/articles/logfire-and-agentsh) — 2026-03-25 ([markdown](https://pydantic.dev/articles/logfire-and-agentsh.md))
  Combine Pydantic Logfire and AgentSH for end-to-end agent observability: trace what the agent thought and audit what it actually did on the machine, in one OpenTelemetry timeline.
- [Reducing Pydantic's memory footprint using bitsets](https://pydantic.dev/articles/pydantic-bitset-performance) — 2026-03-24 ([markdown](https://pydantic.dev/articles/pydantic-bitset-performance.md))
  How Pydantic replaced Python sets with Rust-powered bitsets to cut model instance memory by 55% and speed up validation
- [Pydantic AI Gateway is Moving to Logfire](https://pydantic.dev/articles/gateway-merging-into-logfire) — 2026-03-23 ([markdown](https://pydantic.dev/articles/gateway-merging-into-logfire.md))
  The Pydantic AI Gateway is consolidating into Logfire. Here's why that's a good thing.
- [Debugging Python memory issues in production with memray and AI](https://pydantic.dev/articles/debugging-memory-with-memray-and-ai) — 2026-03-20 ([markdown](https://pydantic.dev/articles/debugging-memory-with-memray-and-ai.md))
  How we used memray to attach to a running Kubernetes pod, found that serializing oversized query results was causing OOM kills, and used Claude Code to trace the fix.
- [OSS in the Age of AI: PyAI Panel discussion](https://pydantic.dev/articles/pyai-oss-panel) — 2026-03-19 ([markdown](https://pydantic.dev/articles/pyai-oss-panel.md))
  Guido van Rossum, Sebastian Ramirez, Jeremiah Lowin, and Samuel Colvin on AI-generated PRs, maintainer burnout, code review in the age of agents, and what companies owe open source.
- [Vstorm's pydantic-deep: Deep Agents for the Pydantic Ecosystem](https://pydantic.dev/articles/pydantic-deep-agents) — 2026-03-18 ([markdown](https://pydantic.dev/articles/pydantic-deep-agents.md))
  pydantic-deep bridges the gap for teams in the Pydantic ecosystem, providing deep agent capabilities: planning, file operations, task delegation, sandboxed execution, built by Vstorm on Pydantic AI.
- [Claude Is Getting Better. What About Us?](https://pydantic.dev/articles/learning-in-the-age-of-ai) — 2026-03-17 ([markdown](https://pydantic.dev/articles/learning-in-the-age-of-ai.md))
  A stranger in my own codebase: on the slow erosion of understanding when AI writes the code, and a small trick that's helping me stay sharp
- [Fighting Fire With Fire: How We're Scaling Open Source Code Review at Pydantic With AI](https://pydantic.dev/articles/scaling-open-source-with-ai) — 2026-03-03 ([markdown](https://pydantic.dev/articles/scaling-open-source-with-ai.md))
  How Pydantic AI lead maintainer turned 4,668 PR review comments into 150 AGENTS.md rules using AI, creating a scalable open source code review workflow with braindump, Claude, and GitHub Actions.
- [Pydantic Monty: you probably don't need a full sandbox](https://pydantic.dev/articles/pydantic-monty) — 2026-02-27 ([markdown](https://pydantic.dev/articles/pydantic-monty.md))
  Introducing Monty, a minimal, secure Python interpreter written in Rust for running AI-generated code. Where it stands and why that matters.
- [Keeping docs in sync with your UI using Claude Code and browser automation](https://pydantic.dev/articles/automating-logfire-docs) — 2026-02-20 ([markdown](https://pydantic.dev/articles/automating-logfire-docs.md))
  How we use Vercel's agent-browser with Claude Code to navigate Pydantic Logfire, capture screenshots, and update documentation in one terminal session.
- [Build Reliable and Observable AI Agents with Pydantic AI and DBOS](https://pydantic.dev/articles/pydantic-ai-dbos) — 2026-02-19 ([markdown](https://pydantic.dev/articles/pydantic-ai-dbos.md))
  How to build fault-tolerant, observable AI agents by combining Pydantic AI with DBOS durable execution and Logfire for end-to-end tracing.
- [The Human-in-the-Loop is Tired](https://pydantic.dev/articles/the-human-in-the-loop-is-tired) — 2026-02-18 ([markdown](https://pydantic.dev/articles/the-human-in-the-loop-is-tired.md))
  On reward functions, dopamine, and what it actually feels like when the code starts writing itself
- [Is your Python web framework really the performance bottleneck?](https://pydantic.dev/articles/web-framework-performance) — 2026-02-17 ([markdown](https://pydantic.dev/articles/web-framework-performance.md))
  Using Logfire to investigate where time is really spent during a web request
- [Pydantic validation just hit 10 billion downloads](https://pydantic.dev/articles/pydantic-validation-10-billion-downloads) — 2026-02-13 ([markdown](https://pydantic.dev/articles/pydantic-validation-10-billion-downloads.md))
  Pydantic, the data validation library for Python, just crossed 10 billion downloads. From a 2017 side project to 550M monthly downloads, here's the story so far.
- [LLM-as-a-Judge: A Practical Guide with Pydantic Evals](https://pydantic.dev/articles/llm-as-a-judge) — 2026-02-11 ([markdown](https://pydantic.dev/articles/llm-as-a-judge.md))
  How to use LLM judges effectively for evaluating AI outputs — the critical distinction between one-size-fits-all and case-specific evaluators, writing good rubrics, and integrating with pydantic-evals.
- [Zero Code Instrumentation with eBPF and Logfire](https://pydantic.dev/articles/zero-code-instrumentation-ebpf-logfire) — 2026-02-10 ([markdown](https://pydantic.dev/articles/zero-code-instrumentation-ebpf-logfire.md))
  A guide to instrumenting applications without code changes using eBPF. Learn how to configure OpenTelemetry eBPF instrumentation to automatically capture HTTP traces and metrics from any application, filter by endpoint, and send telemetry to Logfire.
- [Better Python tests with inline-snapshot](https://pydantic.dev/articles/inline-snapshot) — 2026-02-06 ([markdown](https://pydantic.dev/articles/inline-snapshot.md))
  How Pydantic uses inline-snapshot and dirty-equals for better testing
- [The logs I never read](https://pydantic.dev/articles/the-logs-i-never-read) — 2026-02-05 ([markdown](https://pydantic.dev/articles/the-logs-i-never-read.md))
  Using Logfire MCP with a coding agent to debug a customer issue without ever reading the logs directly.
- [How to build a production agentic app, the Pydantic Way](https://pydantic.dev/articles/building-agentic-application) — 2026-02-03 ([markdown](https://pydantic.dev/articles/building-agentic-application.md))
  A complete guide to building production-ready agentic applications using FastAPI, Pydantic AI, Logfire, and Pydantic Evals
- [Automated Prompt Optimization with GEPA, Pydantic AI, and Pydantic Evals](https://pydantic.dev/articles/prompt-optimization-with-gepa) — 2026-02-02 ([markdown](https://pydantic.dev/articles/prompt-optimization-with-gepa.md))
  Learn how to automate prompt engineering using evolutionary algorithms. Build a complete optimization pipeline with GEPA, pydantic-ai, and pydantic-evals.
- [How shredding JSON is giving Logfire 1000x query speedups](https://pydantic.dev/articles/dynamic-shredding-2026-01-26) — 2026-01-29 ([markdown](https://pydantic.dev/articles/dynamic-shredding-2026-01-26.md))
  How dynamic shredding makes the vast majority of queries in Logfire a thousand times faster, with early adopter customers reporting more than 1000x improvement on query performance.
- [Stop Guessing Why CI Failed: Unifying E2E Tests and Backend Traces with Pydantic Logfire](https://pydantic.dev/articles/tests-observability) — 2026-01-22 ([markdown](https://pydantic.dev/articles/tests-observability.md))
  Connect your Pytest E2E tests to backend traces with Pydantic Logfire. See exactly why tests fail - from assertion to database query - in a single distributed trace.
- [Observability with Logfire and OpenTelemetry Collector](https://pydantic.dev/articles/logfire-opentelemetry-collector) — 2026-01-21 ([markdown](https://pydantic.dev/articles/logfire-opentelemetry-collector.md))
  Learn how to integrate Pydantic Logfire with OpenTelemetry Collector for centralized observability, flexible data routing, and enhanced monitoring capabilities
- [Migrating from Ingress to Gateway API - GKE + Logfire examples](https://pydantic.dev/articles/ingress-to-gateway) — 2026-01-20 ([markdown](https://pydantic.dev/articles/ingress-to-gateway.md))
  A comprehensive guide to migrating from Kubernetes Ingress to Gateway API with zero downtime. Featuring real-world GKE examples, feature mapping tables, and observability best practices with Pydantic Logfire.
- [Logfire MCP: AI-Native Debugging for your Applications](https://pydantic.dev/articles/logfire-mcp-is-awesome) — 2026-01-15 ([markdown](https://pydantic.dev/articles/logfire-mcp-is-awesome.md))
  How to use the Logfire MCP server to give your LLM direct access to live production telemetry, enabling faster and more accurate debugging.
- [Full-stack tracing: Javascript observability with Pydantic Logfire](https://pydantic.dev/articles/javascript-observability) — 2026-01-14 ([markdown](https://pydantic.dev/articles/javascript-observability.md))
  Bridge the gap between your Python backend and JavaScript frontend with Logfire's JavaScript SDK for complete distributed tracing
- [New timeline controls in Logfire](https://pydantic.dev/articles/logfire-timeline-controls) — 2025-12-19 ([markdown](https://pydantic.dev/articles/logfire-timeline-controls.md))
  Pydantic Logfire's timeline now supports precision zooming, flexible time frame adjustment, and intuitive navigation.
- [How Tiger Data Built a Production AI Slack Bot with Pydantic AI and Logfire](https://pydantic.dev/articles/tiger-data-ai-slack-bot-pydantic-logfire) — 2025-12-17 ([markdown](https://pydantic.dev/articles/tiger-data-ai-slack-bot-pydantic-logfire.md))
  Learn how Tiger Data built a production-ready AI Slack bot using Pydantic AI and Logfire with MCP server, distributed tracing, and LLM observability.
- [Pydantic joins the Agentic AI Foundation](https://pydantic.dev/articles/pydantic_agentic_ai) — 2025-12-09 ([markdown](https://pydantic.dev/articles/pydantic_agentic_ai.md))
  Pydantic joins the Linux Foundation's Agentic AI Foundation as a Silver Member to help advance open source standards for agent frameworks.
- [Pydantic Logfire Pricing is Changing](https://pydantic.dev/articles/logfire-pricing-change) — 2025-12-08 ([markdown](https://pydantic.dev/articles/logfire-pricing-change.md))
  Logfire observability pricing changes effective Jan 2026. Free tier: 10M spans/month. New Team ($49) and Growth ($249) plans for scaling teams.
- [Announcement: Launching Pydantic AI Gateway](https://pydantic.dev/articles/pydantic-ai-gateway) — 2025-11-13 ([markdown](https://pydantic.dev/articles/pydantic-ai-gateway.md))
  Pydantic AI Gateway: Manage OpenAI, Anthropic, and other LLMs with one key. Real cost control, OpenTelemetry observability, and failover.
- [Announcement: Pydantic AI support for Vercel AI Elements](https://pydantic.dev/articles/pydantic-ai-ui-vercel-ai) — 2025-11-05 ([markdown](https://pydantic.dev/articles/pydantic-ai-ui-vercel-ai.md))
  Connect Pydantic AI agents to your React app with native streaming support. Build chat interfaces faster with automatic event protocol translation to Vercel AI Elements in Python.
- [Introducing Py AI: Community events for Python AI builders](https://pydantic.dev/articles/pydantic-events) — 2025-10-29 ([markdown](https://pydantic.dev/articles/pydantic-events.md))
  Py AI is a series of events for Python AI developers with industry practitioners about AI framework tools. Events in San Francisco, New York, & London.
- [Pydantic Open Source Fund](https://pydantic.dev/articles/pydantic-oss-fund-2025) — 2025-10-07 ([markdown](https://pydantic.dev/articles/pydantic-oss-fund-2025.md))
  The Pydantic OSS Fund is our commitment to supporting an open source ecosystem: We back open source projects & maintainers that support what we build.
- [Announcement: Pydantic v2.12 Release](https://pydantic.dev/articles/pydantic-v2-12-release) — 2025-10-07 ([markdown](https://pydantic.dev/articles/pydantic-v2-12-release.md))
  What’s new in Pydantic 2.12: Python 3.14 support, experimental MISSING sentinel, TypedDict PEP 728 extras, improved URL path handling & more
- [Pydantic AI v1: A Predictable & Robust GenAI Framework](https://pydantic.dev/articles/pydantic-ai-v1) — 2025-09-04 ([markdown](https://pydantic.dev/articles/pydantic-ai-v1.md))
  Built by engineers, for engineers: Explore dependency injection, type safe, structured outputs, retries & observability for production-ready AI agents
- [Introducing Pydantic’s New Brand Identity](https://pydantic.dev/articles/pydantic-rebranding) — 2025-08-29 ([markdown](https://pydantic.dev/articles/pydantic-rebranding.md))
  We are refreshing our brand to reflect our evolution: Transforming into a unified platform for data validation, AI Agents, observability & more
- [Debug Rust code with OpenAI Codex & Pydantic Logfire](https://pydantic.dev/articles/openai-codex-mcp-talk) — 2025-08-14 ([markdown](https://pydantic.dev/articles/openai-codex-mcp-talk.md))
  Pydantic's David Hewitt dives into Pydantic Logfire, explains how to use the Logfire Rust SDK to instrument OpenAI Codex & introduces Logfire MCP Server
- [How to Integrate Hugging Face Inference in Pydantic AI](https://pydantic.dev/articles/hugging-face-inference-providers-in-pydantic-ai) — 2025-08-05 ([markdown](https://pydantic.dev/articles/hugging-face-inference-providers-in-pydantic-ai.md))
  Step-by-step guide for integrating Hugging Face interference providers with Pydantic AI: Configuration, data model selection, agent setup & more
- [Pydantic Logfire Updates: Dashboards, MCP, Evals & more](https://pydantic.dev/articles/q2-2025-summary) — 2025-07-02 ([markdown](https://pydantic.dev/articles/q2-2025-summary.md))
  Q2 2025 Highlights: Rebuilt dashboard engine, Logfire MCP server integration, improved Pydantic Evals Integration with Logfire & more
- [How to Built an Elite Data Team Leveraging Pydantic AI](https://pydantic.dev/articles/building-data-team-with-pydanticai) — 2025-05-28 ([markdown](https://pydantic.dev/articles/building-data-team-with-pydanticai.md))
  Lessons from building a data team running Pydantic AI: How to set up an AI agent, create integrations & benchmark for fast delivery & reliable results
- [Pydantic Evals Launched: Open-source evals for AI models](https://pydantic.dev/articles/q1-2025-summary) — 2025-04-02 ([markdown](https://pydantic.dev/articles/q1-2025-summary.md))
  Q1 2025 Highlights: Introducing Pydantic Evals, Pydantic Logfire MCP Server and other enhancements for the Pydantic AI platform
- [Announcement: Pydantic v2.11 Release](https://pydantic.dev/articles/pydantic-v2-11-release) — 2025-03-27 ([markdown](https://pydantic.dev/articles/pydantic-v2-11-release.md))
  What’s new in Pydantic 2.11: Faster parsing, reduced memory usage for Pydantic model schemas, extended typing features, and other enhancements
- [Pydantic Logfire Now Available in the EU Region](https://pydantic.dev/articles/eu-region-launch) — 2025-03-26 ([markdown](https://pydantic.dev/articles/eu-region-launch.md))
  Pydantic Logfire adds EU data residency options. Storing data within the European Union enables businesses to meet regional compliance requirements
- [Announcement: Pydantic Logfire JavaScript SDK](https://pydantic.dev/articles/js-sdk-launch) — 2025-03-24 ([markdown](https://pydantic.dev/articles/js-sdk-launch.md))
  We are excited to officially announce the launch of the JavaScript SDK for our observability platform Pydantic Logfire
- [Announcement: Logfire Cloud & Self-Hosted Versions](https://pydantic.dev/articles/logfire-self-hosting-announcement) — 2025-03-21 ([markdown](https://pydantic.dev/articles/logfire-self-hosting-announcement.md))
  We are excited to officially announce Enterprise Cloud and Enterprise Self-Hosted versions of our observability platform Pydantic Logfire
- [Pydantic AI MCP Support & Logfire MCP Server Release](https://pydantic.dev/articles/mcp-launch) — 2025-03-20 ([markdown](https://pydantic.dev/articles/mcp-launch.md))
  We are excited to officially announce the launch of Pydantic AI Agents as MCP clients and the Pydantic Logfire MCP server
- [Announcement: Pydantic Logfire Rust SDK](https://pydantic.dev/articles/rust-sdk-launch) — 2025-03-19 ([markdown](https://pydantic.dev/articles/rust-sdk-launch.md))
  We are excited to officially announce the launch of the Rust SDK for our observability platform Pydantic Logfire
- [Pydantic is Hiring Again](https://pydantic.dev/articles/hiring-again) — 2025-03-17 ([markdown](https://pydantic.dev/articles/hiring-again.md))
  The Pydantic team has new open roles
- [Announcement: Pydantic v2.10 Release](https://pydantic.dev/articles/pydantic-v2-10-release) — 2024-11-13 ([markdown](https://pydantic.dev/articles/pydantic-v2-10-release.md))
  What’s new in Pydantic 2.10: Support for partial validation, default factories taking validated data as an argument and other enhancements
- [How Hyperlint Cut Review Time by 80% with Logfire](https://pydantic.dev/articles/why-hyperlint-chose-logfire-for-observability) — 2024-10-10 ([markdown](https://pydantic.dev/articles/why-hyperlint-chose-logfire-for-observability.md))
  Hyperlint chose Pydantic Logfire to monitor AI workflows, trace spans and to integrate OpenAI & Anthropic. Learn why it turned out to be a true game-changer
- [The Pydantic Open Source Fund](https://pydantic.dev/articles/pydantic-oss-fund-2024) — 2024-10-07 ([markdown](https://pydantic.dev/articles/pydantic-oss-fund-2024.md))
  The Pydantic OSS Fund is our commitment to supporting an open source ecosystem: We back open source projects & maintainers. Eligibility, awards & timelines
- [Why is Pydantic building an Observability Platform?](https://pydantic.dev/articles/why-logfire) — 2024-10-01 ([markdown](https://pydantic.dev/articles/why-logfire.md))
  Reasons why the team behind the Pydantic Validation library is building an observability platform
- [Announcement: Pydantic Logfire launch & Series A Funding](https://pydantic.dev/articles/logfire-announcement) — 2024-10-01 ([markdown](https://pydantic.dev/articles/logfire-announcement.md))
  Q4 2024 Highlights: Pydantic Logfire is leaving beta and Series A funding led by Sequoia with participation from Partech and Irregular Expression securedg
- [Announcement: Pydantic v2.9 Release](https://pydantic.dev/articles/pydantic-v2-9-release) — 2024-09-05 ([markdown](https://pydantic.dev/articles/pydantic-v2-9-release.md))
  What’s new in Pydantic 2.9: Complex number support, explicit ZoneInfo support, val_json_bytes setting, and other enhancements
- [Announcement: Pydantic v2.8 Release](https://pydantic.dev/articles/pydantic-v2-8-release) — 2024-07-01 ([markdown](https://pydantic.dev/articles/pydantic-v2-8-release.md))
  What’s new in Pydantic 2.8: Fail-fast validation, extended model/field deprecation in JSON schema, programmatic titles & serialization context for TypeAdapter
- [Announcement: Pydantic v2.7 Release](https://pydantic.dev/articles/pydantic-v2-7-release) — 2024-04-11 ([markdown](https://pydantic.dev/articles/pydantic-v2-7-release.md))
  What’s new in Pydantic 2.7: Partial JSON parsing, generic Secret types, duck-typing serialization as an opt-in feature, plus performance Improvements
- [AWS Lambda: Validate event & context data via Pydantic](https://pydantic.dev/articles/lambda-intro) — 2024-04-04 ([markdown](https://pydantic.dev/articles/lambda-intro.md))
  Practical guide with step-by-step instructions for validating event and context data in AWS Lambda functions using Pydantic
- [Building a product search API with GPT-4 Vision, Pydantic, and FastAPI](https://pydantic.dev/articles/llm-vision) — 2024-02-29 ([markdown](https://pydantic.dev/articles/llm-vision.md))
  Tutorial on how to use Pydantic with GPT-4 Vision & FastAPI to process user provided images, validate results & power a product search API for ecommerce apps
- [Ho to Minimize LLM Hallucinations with Pydantic Validators](https://pydantic.dev/articles/llm-validation) — 2024-01-18 ([markdown](https://pydantic.dev/articles/llm-validation.md))
  This guide shows how you can plug Pydantic validation into your AI pipeline to assert constraints on LLM models, use context to reduce errors & hallucinations
- [How to Use Pydantic for LLMs: Schema, Validation & Prompts](https://pydantic.dev/articles/llm-intro) — 2024-01-04 ([markdown](https://pydantic.dev/articles/llm-intro.md))
  Practical guide on how to use Pydantic to work with large language models (LLMs): Define schemas and generate structured JSON, type-safe responses
- [Announcement: Pydantic V2 Release](https://pydantic.dev/articles/pydantic-v2-final) — 2023-06-30 ([markdown](https://pydantic.dev/articles/pydantic-v2-final.md))
  Official release note for Pydantic V2 - with migration guide
- [Help Us Build Our Roadmap](https://pydantic.dev/articles/roadmap) — 2023-06-13 ([markdown](https://pydantic.dev/articles/roadmap.md))
  Pydantic shares its product roadmap and invites community feedback on the developer tools and services being built around the Pydantic ecosystem.
- [Announcement: Pydantic v2 Pre Release](https://pydantic.dev/articles/pydantic-v2-alpha) — 2023-04-03 ([markdown](https://pydantic.dev/articles/pydantic-v2-alpha.md))
  What’s new in Pydantic v2: 5-50x faster than Pydantic v1 and with substantial architecture improvements for enhanced safety & maintainability
- [Company Announcement | Pydantic](https://pydantic.dev/articles/company-announcement) — 2023-02-16 ([markdown](https://pydantic.dev/articles/company-announcement.md))
  Pydantic announces its incorporation and seed funding led by Sequoia Capital to build developer tools and cloud services around the Pydantic ecosystem.
- [Introducing Pydantic v2 - Key Features](https://pydantic.dev/articles/pydantic-v2) — 2022-07-10 ([markdown](https://pydantic.dev/articles/pydantic-v2.md))
  Pydantic v2 reshapes validation with better performance, strict mode, consistency around data conversions & more. New features & migration steps explained