The Curious Engineer

All blogs / AI / ML

Martin Alderson

martinalderson.com

Web development, AI tooling, and building better software

AI / MLSoftware / engineering practice

’25’26
Posts
72 from 10 Aug 2025 to 29 Oct 2026
Posting
weekly (60 in the last year)
Kind
personal blog
Feed
https://martinalderson.com/feed.xml

2026 50

  1. Making agents fast - what I've learnt over the last 12 months
  2. Interesting systemd features you may not know about
  3. How the AI boom might really end: 5 scenarios
  4. Shopify went back to native. I think the bigger shift is formal verification
  5. The AI margin collapse is gathering pace
  6. Have the frontier labs mixed up AI safety and security?
  7. Reducing codebase cognitive debt through... quizzes?
  8. What GLM-5.3 Flash running on Chinese hardware actually means
  9. The summer of open weights
  10. How I think about reducing AI costs
  11. Watch out for cache read costs
  12. I'm (mostly) picking models on speed now, not intelligence
  13. The first known runaway AI agent - or a very bad marketing stunt?
  14. Winners and losers in the coming AI margin collapse (part 2)
  15. GLM 5.2 and the coming AI margin collapse (part 1)
  16. Expert-aware quantisation: near-Q4 quality at near-Q2 size?
  17. A brief history of KV cache compression developments
  18. xAI is looking more like a datacentre REIT than a frontier lab
  19. Is datacentre sovereignty really that important?
  20. I went on the Built for Turbulence podcast
  21. What's going on with Gemini?
  22. Managed agents are the new Lambda
  23. Open weights are quietly closing up - and that's a problem
  24. 29th August 2026: a scenario
  25. Local LLM speed calculator A back-of-envelope tokens/sec estimator for running LLMs locally. Pick your memory, model size and quantization to see how fast it'll actually run.
  26. Figma's woes compound with Claude Design
  27. A little tool to visualise MoE expert routing
  28. Has Mythos just broken the deal that kept the internet safe?
  29. What next for the compute crunch?
  30. Telnyx, LiteLLM and Axios: the supply chain crisis
  31. Using agents and Wine to move off Windows
  32. Why Claude's new 1M context length is a big deal
  33. How to use the Qwen 3.5 LLMs to OCR documents
  34. No, it doesn't cost Anthropic $5k per Claude Code user
  35. Is the AI Compute Crunch Here?
  36. Why on-device agentic AI can't keep up
  37. Using OpenCode in CI/CD for AI pull request reviews
  38. Which web frameworks are most token-efficient for AI agents?
  39. How to convert Excel spreadsheets to Python models with Claude Code A step-by-step guide to converting complex Excel financial models into Python using Claude Code, enabling faster scenario testing and integration with other tools
  40. Who fixes the zero-days AI finds in abandoned software?
  41. Attack of the SaaS clones
  42. How to generate good looking reports with Claude Code, Cowork or Codex
  43. Self-improving CLAUDE.md files
  44. Wall Street just lost $285 billion because of 13 markdown files
  45. Two kinds of AI users are emerging. The gap between them is astonishing.
  46. Turns out I was wrong about TDD
  47. Why sandboxing coding agents is harder than you think
  48. The Coming AI Compute Crunch
  49. Which programming languages are most token-efficient?
  50. I ported Photoshop 1.0 to C# in 30 minutes

2025 22

  1. Why I'm building my own CLIs for agents
  2. Travel agents took 10 years to collapse. Developers are 3 years in.
  3. Are we dismissing AI spend before the 6x lands?
  4. Minification isn't obfuscation - Claude Code proves it
  5. AI agents are starting to eat SaaS
  6. Has the cost of building software just dropped 90%?
  7. Are we in a GPT-4-style leap that evals can't see?
  8. I Finally Found a Use for IPv6
  9. How I use Claude Code to manage sysadmin tasks
  10. Could Excel agents unlock $1T in economic value?
  11. Are we really repeating the telecoms crash with AI datacenters?
  12. A non-technical CFO is shipping better code than the agencies he hired
  13. Tracking MCP Server Growth
  14. Notes from MCP Dev Summit Europe: Where the Protocol Is Headed
  15. How I make CI/CD (much) faster and cheaper
  16. Google AI Studio API has been unreliable for the past 2 weeks
  17. What happens when coding agents stop feeling like dialup?
  18. Solving Claude Code's API Blindness with Static Analysis Tools
  19. Are OpenAI and Anthropic Really Losing Money on Inference?
  20. I gave Claude Code a folder of tax documents and used it as a professional tax agent
  21. Beyond the Hype: Real-World MCP Support Across Major AI APIs
  22. Welcome to My Blog