AI, tested in the real world
AI news, agents, automation and practical experiments focused on what actually works.
SIGNAL → TEST → USE
FLOW_01 // 03
What’s happening. What I test. What it means in practice.
01 // SIGNAL
What’s happening in AI and why it matters.
02 // TEST
I try the tools, agents and workflows myself.
03 // USE
What actually works, and where it can help a business.
// LIVE SIGNAL FEED
LATEST SIGNAL
SIG_01
The AI stories worth paying attention to, with a practical view on why they matter.
SIGNAL · 13 Oct 2026
Muse Wonder puts Meta back at the AI frontier
Meta’s Muse Wonder Full has moved the company into the leading group of frontier models, with independent testing placing it close to the strongest systems from Anthropic, OpenAI and Google.[2] The model matters, but Meta’s larger advantage may be the infrastructure and distribution already assembled around it.[1][3]
SIGNAL · 07 Oct 2026
When mathematical discovery starts scaling with compute
OpenAI has released 722 AI-produced mathematics manuscripts spanning 372 groups of related results.[2] The volume hints at scientific discovery scaling with compute, but the uneven verification of the work shows why mathematics is only the beginning of the test.[3]
// ACTIVE LAB
FROM THE LAB
LAB_02
Hands-on experiments with AI, agents, automation and the tools I’m actually using.
LAB
Building an AI Content Workflow Without Giving Up Editorial Control
How I built an AI content workflow around human editorial judgement rather than replacing it with automation.
WHAT I TESTED
Building an AI-assisted editorial workflow for DEVKEV that automates discovery, research support, review and publishing mechanics while keeping story selection, editorial judgement and publication approval under my control.
VERDICT
The strongest version of the system is not the most autonomous one. Automation works best around the editorial judgement rather than replacing it, with the final decision about what deserves attention and what becomes public still remaining mine.
Read the Lab note → View all Lab →CURRENTLY TESTING
● ACTIVE
The tools, workflows and experiments I’m actively trying right now.
Testing
Elementor MCP through Hermes
Testing whether Hermes can connect to and work with a real Elementor site through MCP, while monitoring permissions, resource use and the quality of its decisions.
STARTED
14/10/2026
LATEST NOTE
The initial setup has already exposed some useful friction. Wordfence disables WordPress Application Passwords by default, so I had to make a deliberate security exception before Elementor could generate the connection.I created a separate WordPress administrator account for Rez, protected its dashboard login with 2FA, and generated the MCP setup through Elementor’s generic “Other” option. I have given Rez the original setup prompt unchanged to see whether it understands the instructions, adapts them to Hermes and establishes the connection correctly.No changes to the site have been authorised yet.
Working
AI CONTENT WORKFLOW
STARTED
09/09/2026
LATEST NOTE
Signal is now running through n8n, Discord and Rez/Hermes, with manual story selection and final approval retained. I’m watching how it behaves with real stories before automating distribution further.
WORK WITH ME
If you’re trying to work out where AI, agents or automation could genuinely help your business, I can help you test the options and build something practical
Tell me what you're trying to do