The Curious Engineer

All blogs / AI / ML

Anup Jadhav

anup.io

I build AI agent systems that hold up in production. My focus is the agent harness: orchestration, durable workflows and observability.

AI / MLSoftware / engineering practice

’19’20’21’22’23’24’25’26
Posts
109 from 13 Jul 2019 to 19 Oct 2026
Posting
weekly (49 in the last year)
Kind
personal blog
Feed
https://www.anup.io/rss/

2026 40

  1. The Next Wave of AI Engineering Is About Everything Around the Model
  2. What is Jev? A Simple Primer on AI That Makes Decisions
  3. TIL: The AI Wall
  4. Defining an AI Agent
  5. TIL: What’s actually different about GGUF and ONNX?
  6. How I code with Claude Code
  7. One agent or many?
  8. From Prompt Engineering to Loop Engineering
  9. Who Still Understands the Code?
  10. Designing teams for an agentic world
  11. The Frontier of Agent Memory: From Recall to Experience
  12. Speculative Decoding
  13. Query, Key, Values
  14. How Modern Agent Memory Architectures Work
  15. Why Context Is Not Enough
  16. On Durable Objects, Orleans, and prior art for the agentic web
  17. TIL: Ads in AI chatbots are not just a UX problem
  18. Welcome to Middle Loop Engineering
  19. How fast does it serve? Throughput, latency, and picking the right GPU
  20. Fitting LLMs on Self-Hosted GPUs
  21. The Harness Is the Product
  22. How "Thinking" Models Actually Work
  23. Harness Engineering by Anup Jadhav
  24. We’re Being Too Loose With the Term “World Model”
  25. TIL: Quantisation
  26. Write Skills Like Workstations, Not Prompts
  27. How the Claude Code team designs agent tools
  28. Make Claude Code Review Its Own Plans
  29. Ship Prompts Like Software: Regression Testing for LLMs
  30. Four Ways to Grade an LLM (Without Going Broke)
  31. Your Golden Dataset Is Worth More Than Your Prompts
  32. Build LLM Evals You Can Trust
  33. Stripe's coding agents: the walls matter more than the model
  34. Deep Blue
  35. TIL: The real bottleneck in AI coding isn't speed
  36. TIL: Markov Language
  37. Become an AI Value Creator
  38. Claude Code Tips From the Guy Who Built It
  39. TIL: Learning New Tech With AI Assistance Might Backfire
  40. Temporal + LangGraph: A Two-Layer Architecture for Multi-Agent Coordination

2025 61

  1. AI's Naughty and Nice List
  2. Why AI Coding Advice Contradicts Itself
  3. The Context Advantage
  4. Engineering teams evolve from coders to orchestrators
  5. What It Takes To Serve 10,000 Queries A Day | Anup Jadhav
  6. Understanding Generative UI
  7. LLMs develop distinct trading personalities when given real money
  8. Senior Developers Are Better at Using AI Agents |Anup Jadhav
  9. pgvector doesn't scale
  10. TIL: How Transformers work
  11. Fine-Tuning makes a comeback
  12. How to read research papers
  13. Eliza Redux: A Real-Time Voice AI Crisis Support Agent
  14. Atlas Unshrugged
  15. Choosing an LLM Is Choosing a World-View
  16. Reasoning LLMs are wanderers rather than systematic explorers
  17. Stop Planning Your AI Strategy and Start Running Experiments
  18. Introducing AI Engineering Premium
  19. Horizon Length - Moore's law for AI Agents
  20. The boring secret to building better AI agents
  21. Representation Engineering with Control Vectors
  22. Rethinking RAG: Meta’s REFRAG
  23. Nobody Knows How to Charge for AI
  24. Pace Layering Framework
  25. MCP + A2A: The Protocols Making AI Agents Actually Work Together
  26. AI Agents Are Finally Fixing Real-World Code Security Problems
  27. Data in CSV Format Isn’t Always the Best for LLMs
  28. Prompt Engineering vs Context Engineering
  29. Why LLMs Confidently Hallucinate a Seahorse Emoji That Never Existed
  30. Goodbye Manual Prompts, Hello DSPy
  31. Nemawashi (根回し)
  32. AI agents are starting to do real work
  33. Cargo Cult Vibe Coding
  34. The Rubber Duck That Talks Back
  35. Why Your AI Agents Keep Breaking (And How Systems Thinking Can Fix Them)
  36. From CI/CD to CAI (Continuous AI)
  37. The Agentic Protocols That Will Define the Next Decade of Software
  38. When Regulators Embrace AI
  39. AI Agents Are Getting Twice as Capable Every 7 Months
  40. Guardrails for Agentic AI
  41. The End of Thinking (Again)
  42. Five Hypothesis On The Future of AI
  43. Vibe Coding, Revisited: From Weekend Chaos to Repeatable Craft
  44. Protocols for the Agentic Era
  45. Why AI Safety Is Harder Than It Sounds: Lessons from DeepMind’s AGI Strategy
  46. Architecting the Internet of Agents
  47. 10 Ways to Vibe Code
  48. Streamlining AI-to-API Communication: A Practical Guide to Model Context Protocol
  49. From Coding to Spec-Writing - Navigating AI's impact on Devs
  50. Choosing Between RAG, Fine-Tuning, or Hybrid Approaches for LLMs
  51. RAGEntic AI: A framework for AI Agents
  52. LLM Security 101: Defending Against Prompt Hacks
  53. The Art of Sampling: Controlling Randomness in LLMs
  54. Deep Dive into LLMs like ChatGPT by Andrej Karpathy - TL;DW version
  55. Thinking Smarter, Not Harder: How LLMs Can Learn on the Fly
  56. How to think about LLM Model Size
  57. Building, Breaking, and Rebuilding: My Journey to a Smarter RAG Application
  58. From ML to AI Engineering: Transforming How We Build AI Applications
  59. Building a Matrix-themed Chat Interface with Heroku's New Managed Inference API
  60. AI Agent Use Case Evaluation: From Risk Assessment to Implementation
  61. LLM Transformer Overview

2024 2

  1. What is an AI Agent?
  2. LLMCompiler - Towards parallel function calling

2022 1

  1. Anybody can learn to code

2021 2

  1. How I learned to stop worrying and love SecOps
  2. Invert, always, invert

2020 1

  1. Tracking Productivity

2019 2

  1. Documentation - your secret weapon
  2. From monolith to microservices - a journey