
What is the best AI classifier? I made them fight to the death to find out.
AI classifiers battle in Classifier Death Match. Explore 120 matches, latency and decision quality, and what changes when models choose their own targets.
Posts from Inkie Marketing Platform.

AI classifiers battle in Classifier Death Match. Explore 120 matches, latency and decision quality, and what changes when models choose their own targets.

A faster TypeScript compiler could remove hours of waiting across our AI coding agents. Real tsgo, tsc-ts and classic tsc timings, with the caveats that matter.

Discover how Claude Code's support for AGENTS.md and omitClaudeMd helps optimise token spend and standardise instruction rules across AI coding agents.

Discover how benchmarking Jev against Gemini for coding agent safety gates reveals vital insights into speed, probability thresholds and AI decision logic.

Discover what happens when you scale up parallel AI agents and experience agentic cognitive overload. Learn how to manage AI workflow burnout effectively.

Why one live timestamp in a cached system prompt busts your entire prompt cache, and the one-line fix that stops it happening.

A three-layer security classifier that lets autonomous agents act without constant approval, built runtime-agnostic from day one.

Confused by MCP and A2A? Discover the key differences between agent-to-service and agent-to-agent protocols, and find out which one you actually need today.

Why a second opinion only works if it comes from a genuinely different model, and how Antigravity reviews code Claude Code already approved.

Sophie Boulderstone published UAIL, a 0 to 10 scale for real AI capability. Simon Dixon on why he open sourced it on GitHub under CC BY 4.0 the same day.

The Brain: Simon's long-term memory system for AI agents, markdown files, decision cards, vector search, and sidecars, explained end to end.

How to make a scarce model's usage last: Fable plans and verifies, Sonnet or Opus does the building, same quality, far fewer tokens burned.

MCP or CLI for your agent's tools: it comes down to where the agent runs, your own machine or a server you don't fully control.

MCP or a raw API for agent tooling: one interface you learn once and reuse everywhere, versus one you rebuild for every service.

Discover how AI is democratising custom software for small businesses. Learn how to build bespoke tools tailored to your unique workflows without costly fees.

Discover why small, consistent steps beat grand marketing plans. Learn how to adopt the maker's mindset to build trust and grow your business today.

Tired of fighting rigid software? Learn how to design your tech stack around your unique lifestyle and workflow to reclaim your creative freedom today.

How memory actually gets into Claude Code: the CLAUDE.md trap, the SessionStart hook fix, and the real 10K-token hook limit.

AI agents forget past a certain point. Four ways to fix it: memory files, RAG, the file system, and observational memory, with the real tradeoffs of each.

Join Simon Dixon at EVOLVE Brighton today as he live-demos how Inkie’s structured multi-agent AI pipeline automates small business marketing. 50% ticket discount inside!