Writing, tools, & open source.

• May 2026 How to use AI effectively - the missing manual
  Pick the best model you have, invest in the first prompt, debug by fixing the prompt and re-running. Practical techniques: precision and self-consistency, explicit tradeoff resolution, instructing the model to browse, top-down walkthrough to locate the first divergence, decomposition. X
• Jan 2026 The Best Harness Beats the Best Model
  The code+prompt layer matters more than the model in 2026. Small and open models in a good harness rival frontier models for most tasks. Concrete ordering for general use and for coding, with notes on where local-first works today. X
• Jan 2026 Can we trust Claude Code even more?
  Trust is the bottleneck that keeps users from getting senior-level performance from AI coding assistants. Proposes a graduated trust model (junior → mid → senior → yolo) tuned per-user, with a thumbs-up/down scorecard per stage. X
• Dec 2025 Building ahai - what, how, why
  A transparent decision log for ahai. Local-first MLX, Mac-only, $19+ one-time pricing, no subscriptions, MLX over Ollama (own benchmarks), US-origin default model, PySide6 over alternatives, no DRM. Meta-thesis: "open source / free as bait" is morally worse than transparent paid. X
• Nov 2025 My AI coding mindset
  A compact manifesto on AI-assisted coding. End-to-end ownership, recursive task decomposition, per-task AI-vs-self decisions, tests as immutable contracts, commit often (even broken - labeled), dual docs, parallel sessions designed for diversity. X
• Nov 2025 IRS deploys AI Agents
  Brief commentary on AI agents being deployed at IRS scale and the implications for tax processing and human-software interfaces. X
• Jul 2025 The LLM Paradox
  Humans are conditioned to trust convincing natural language. LLMs are trained to sound more convincing than most humans, yet still err. The mismatch creates instinctive trust, repeated letdown, and a subconscious addiction to the conversational form. X
• Jun 2025 LLM Filter Bubble
  Beyond well-known geo and ideological training biases: a structural economic bias. Proprietary frontier labs have little incentive to surface high-quality research about open and local models that would threaten their business. X

Latest Posts

• Mar 2025 My Goto Web Chat AI Agents Bear Blog
• Mar 2025 An Ode to Claude in Cursor Bear Blog
• Mar 2025 A Cheap Way to Compare LLM Knowledge Bear Blog
• Mar 2025 Long Conversations with LLMs via Language-as-Code Bear Blog

Hashnode Highlights

• Dec 2024 State of AI End of 2024: A Brain Dump 20 min
• Feb 2025 List of Web Chat AI Models 2 min
• Jan 2025 A Penny for DeepSeek-R1's Thoughts 4 min
• Jan 2025 Evaluating Local Models with Custom Datasets 5 min
• Jan 2025 DeepSeek-R1 Thinks in Cursor 4 min
• Jan 2025 AI Disruptors 3 min
• Sep 2024 Key Components of My AI and Technology Stack 3 min
• Aug 2024 What Could Reduce Current AI Demand for GPUs 3 min
• Jul 2024 AI Assisted Software Development 5 min
• Jun 2024 Managing AI Anxiety 24 min

Medium Highlights

• Sep 2024 Rethinking Structured Data in a World with AI - pinned Medium
• Sep 2024 Switch to ChatGPT o1 - pinned Medium
• Sep 2024 Using Chat AI - pinned Medium
• Sep 2024 Fallacy of Shared Identities - pinned Medium
• Jan 2025 Evaluating Software by its Config Medium

Public Repositories

170+ repos on GitHub (mostly private during active development).

Hacker News - rcanand2025

• Jan 2026 Show HN: Visualize full Claude Code CLI sessions in a rich UI Show HN
• Dec 2025 Show HN: Visualize different text streaming speeds Show HN
• Dec 2025 Is Vibe Coding Safe Story
• Dec 2025 Show HN: Ahai - Find your ideas scattered across files Show HN

Quora - Anand C Ramanathan

50+ answers with 1M+ views and 1K+ upvotes across technology, AI, and software development topics.