Field Notes — FieldprovenPractitioner notes on AI where performance marketing meets data: what is genuinely shifting, and what we learn building in it.https://fieldproven.ai/enAn AI agent will hand you an analysis that is confidently, beautifully wronghttps://fieldproven.ai/field-notes/looking-correct-vs-being-correct/https://fieldproven.ai/field-notes/looking-correct-vs-being-correct/Looking correct and being correct are different things in data work. Where agent-built analyses quietly break, and what the expert is still there for.Thu, 04 Jun 2026 00:00:00 GMTPerformance marketing × data, on AIandy@fieldproven.ai (Andy Goldschmidt)You don't need to AI-automate all of performance marketinghttps://fieldproven.ai/field-notes/ai-quick-wins-performance-marketing/https://fieldproven.ai/field-notes/ai-quick-wins-performance-marketing/Four AI quick wins that move a number in paid media without betting the account: negative keywords, ad copy angles, creative variants, ad-hoc analysis.Tue, 19 May 2026 00:00:00 GMTPerformance marketing × data, on AIandy@fieldproven.ai (Andy Goldschmidt)Where the LLM wiki pattern actually fitshttps://fieldproven.ai/field-notes/where-the-llm-wiki-fits/https://fieldproven.ai/field-notes/where-the-llm-wiki-fits/The LLM-maintained wiki works at one-operator scale. Two cracks before you build it for a team: the input bar, and knowledge that will not hold still.Tue, 12 May 2026 00:00:00 GMTMaking AI work in your teamandy@fieldproven.ai (Andy Goldschmidt)LLMs might be better coders, but humans are better communicatorshttps://fieldproven.ai/field-notes/llms-better-coders-humans-better-communicators/https://fieldproven.ai/field-notes/llms-better-coders-humans-better-communicators/AI does not have an output problem, it has an input problem. Why technically correct but practically wrong results are a specification failure.Wed, 06 May 2026 00:00:00 GMTMaking AI work in your teamandy@fieldproven.ai (Andy Goldschmidt)Claude Dispatch & Channelshttps://fieldproven.ai/field-notes/claude-dispatch-and-channels/https://fieldproven.ai/field-notes/claude-dispatch-and-channels/Dispatch and Channels put my Claude Code setup on my phone — how they compare with OpenClaw, and the one tradeoff both of them still require.Wed, 01 Apr 2026 00:00:00 GMTBuilding with AIandy@fieldproven.ai (Andy Goldschmidt)The Bullshit Benchmarkhttps://fieldproven.ai/field-notes/the-bullshit-benchmark/https://fieldproven.ai/field-notes/the-bullshit-benchmark/Models ace coding challenges, but will they push back when the question itself is broken? What BullshitBench measures, and why the scores should worry you.Thu, 19 Mar 2026 00:00:00 GMTPerformance marketing × data, on AIandy@fieldproven.ai (Andy Goldschmidt)What I actually gained switching to Claudehttps://fieldproven.ai/field-notes/what-i-actually-gained-switching-to-claude/https://fieldproven.ai/field-notes/what-i-actually-gained-switching-to-claude/The difference is not features but whether the AI assists your work or does it — on agentic harnesses, and a model that pushes back on weak assumptions.Thu, 12 Mar 2026 00:00:00 GMTThe AI landscape, beyond the hypeandy@fieldproven.ai (Andy Goldschmidt)What I actually lost switching to Claudehttps://fieldproven.ai/field-notes/what-i-actually-lost-switching-to-claude/https://fieldproven.ai/field-notes/what-i-actually-lost-switching-to-claude/The honest cost of leaving ChatGPT for Claude: price, voice mode, image generation, and availability — the tradeoffs the comparison posts leave out.Wed, 04 Mar 2026 00:00:00 GMTThe AI landscape, beyond the hypeandy@fieldproven.ai (Andy Goldschmidt)Non-determinism is not a problemhttps://fieldproven.ai/field-notes/non-determinism-is-not-a-problem/https://fieldproven.ai/field-notes/non-determinism-is-not-a-problem/Humans give different answers on different days too. Judge AI on the quality of its solutions, not on the consistency of the path it took to get there.Wed, 10 Dec 2025 00:00:00 GMTThe AI landscape, beyond the hypeandy@fieldproven.ai (Andy Goldschmidt)Recap MAS Berlin 2025https://fieldproven.ai/field-notes/recap-mas-berlin-2025/https://fieldproven.ai/field-notes/recap-mas-berlin-2025/Notes from speaking at Marketing Analytics Summit Berlin: AI adoption cannot be templated, clean data still decides, and human judgment gains value.Tue, 18 Nov 2025 00:00:00 GMTThe AI landscape, beyond the hypeandy@fieldproven.ai (Andy Goldschmidt)Context beats prompthttps://fieldproven.ai/field-notes/context-beats-prompt/https://fieldproven.ai/field-notes/context-beats-prompt/AI projects fail on missing business context, not model intelligence. Build reusable context layers — goal, audience, boundaries, metrics — and reuse them.Thu, 16 Oct 2025 00:00:00 GMTMaking AI work in your teamandy@fieldproven.ai (Andy Goldschmidt)Claude Code as My Second Brainhttps://fieldproven.ai/field-notes/claude-code-as-my-second-brain/https://fieldproven.ai/field-notes/claude-code-as-my-second-brain/How a coding agent became a command center for project management and strategic work: CLAUDE.md, MCP tools, sub-agents, and custom slash commands.Wed, 01 Oct 2025 00:00:00 GMTBuilding with AIandy@fieldproven.ai (Andy Goldschmidt)GPT-5 reactionhttps://fieldproven.ai/field-notes/gpt-5-reaction/https://fieldproven.ai/field-notes/gpt-5-reaction/Not the doom-and-gloom take: where GPT-5-Thinking clearly beats o3 in hands-on data analysis, and why it is an evolution rather than a revolution.Wed, 13 Aug 2025 00:00:00 GMTThe AI landscape, beyond the hypeandy@fieldproven.ai (Andy Goldschmidt)The AI App Store Is Here: How MCPs Just Changed Everythinghttps://fieldproven.ai/field-notes/the-ai-app-store-is-here-how-mcps-just-changed-everything/https://fieldproven.ai/field-notes/the-ai-app-store-is-here-how-mcps-just-changed-everything/MCP took our AI out of the chat window and connected it to the tools the team already uses — what that unlocked, and where it still meets reality.Wed, 23 Jul 2025 00:00:00 GMTThe AI landscape, beyond the hypeandy@fieldproven.ai (Andy Goldschmidt)Claude Code impressionhttps://fieldproven.ai/field-notes/claude-code-impression/https://fieldproven.ai/field-notes/claude-code-impression/A weekend with Claude Code: a production-grade prompt library, a Chrome extension from scratch, and why iterative execution beats snippet generation.Mon, 14 Jul 2025 00:00:00 GMTBuilding with AIandy@fieldproven.ai (Andy Goldschmidt)Vibe Coding: When to Trust the AI, and When to Take the Wheelhttps://fieldproven.ai/field-notes/vibe-coding-when-to-trust-the-ai-and-when-to-take-the-wheel/https://fieldproven.ai/field-notes/vibe-coding-when-to-trust-the-ai-and-when-to-take-the-wheel/Where letting the AI drive pays off and where it costs you: the learning, quality, and auditability pitfalls of vibe coding, and how to pick your line.Wed, 26 Mar 2025 00:00:00 GMTBuilding with AIandy@fieldproven.ai (Andy Goldschmidt)The Learning Paradox: Developing Analytics Talent in the AI Erahttps://fieldproven.ai/field-notes/the-learning-paradox-developing-analytics-talent-in-the-ai-era/https://fieldproven.ai/field-notes/the-learning-paradox-developing-analytics-talent-in-the-ai-era/AI coding tools accelerate analysts and quietly remove the struggle that builds skill. How to keep developing real analytics talent in the AI era.Wed, 26 Feb 2025 00:00:00 GMTMaking AI work in your teamandy@fieldproven.ai (Andy Goldschmidt)How I Built Three Apps Without Writing Code: My Journey with Cursorhttps://fieldproven.ai/field-notes/how-i-built-three-apps-without-writing-code/https://fieldproven.ai/field-notes/how-i-built-three-apps-without-writing-code/Three finished projects in two months after four years without coding, built with Cursor — what the tool does well, and how to get the most out of it.Thu, 13 Feb 2025 00:00:00 GMTBuilding with AIandy@fieldproven.ai (Andy Goldschmidt)Making AI Work in Data Teams: A Practical Guidehttps://fieldproven.ai/field-notes/making-ai-work-in-data-teams-a-practical-guide/https://fieldproven.ai/field-notes/making-ai-work-in-data-teams-a-practical-guide/AI adoption in a data team grows, it does not roll out: low entry points, champions, hands-on workshops, and the expectations that keep it honest.Tue, 21 Jan 2025 00:00:00 GMTMaking AI work in your teamandy@fieldproven.ai (Andy Goldschmidt)Reasoning, Multimodality, and Agents: Three developments every data professional needs to knowhttps://fieldproven.ai/field-notes/reasoning-multimodality-and-agents/https://fieldproven.ai/field-notes/reasoning-multimodality-and-agents/Three developments reshaping marketing and analytics work in 2025 — reasoning models, true multimodality, and autonomous agents — and what each one changes.Thu, 09 Jan 2025 00:00:00 GMTThe AI landscape, beyond the hypeandy@fieldproven.ai (Andy Goldschmidt)AI as Your Management Copilothttps://fieldproven.ai/field-notes/ai-as-your-management-copilot/https://fieldproven.ai/field-notes/ai-as-your-management-copilot/How an analytics manager uses AI beyond dashboards and models: sharper OKRs, faster stakeholder communication, better retrospectives, smarter hiring.Fri, 20 Dec 2024 00:00:00 GMTPerformance marketing × data, on AIandy@fieldproven.ai (Andy Goldschmidt)AI as Your Analytics Coachhttps://fieldproven.ai/field-notes/ai-as-your-analytics-coach/https://fieldproven.ai/field-notes/ai-as-your-analytics-coach/Using AI as a critical thinking partner instead of a generator: dashboard design reviews, SQL query optimization, and stakeholder emails that land.Mon, 02 Dec 2024 00:00:00 GMTPerformance marketing × data, on AIandy@fieldproven.ai (Andy Goldschmidt)Build Analytics Prototypes at the Speed of Conversationhttps://fieldproven.ai/field-notes/build-analytics-prototypes-at-the-speed-of-conversation/https://fieldproven.ai/field-notes/build-analytics-prototypes-at-the-speed-of-conversation/How Claude Artifacts turn analytics prototyping into a conversation: proof-of-concept visuals, quick A/B readouts, flow charts, and where it breaks down.Mon, 18 Nov 2024 00:00:00 GMTBuilding with AIandy@fieldproven.ai (Andy Goldschmidt)