Articles
Longer versions of what I post on LinkedIn, kept here so they don't disappear into the feed.
“Not every AI job needs a model that can write a poem.”
2026-10-02
Jev is a decision-only AI model that returns typed answers with confidence scores. What it is, where it fits in a workflow, and where it doesn't.
“Year one of a new business is a strange kind of broke.”
2026-09-17
No marketing team, no agency retainer, and limited hours. How a brand voice document, a drafting tool, and a human in the loop keep marketing steady in year one.
“The concern is no longer that someone asks a chatbot to write malicious code.”
2026-09-14
Anthropic disrupted Claude misuse across weapons development, cyber operations, and fraud in 2026. Why the shift to AI running whole operations matters.
“The excuse that AI is too expensive just ran out of shelf life.”
2026-09-08
OpenAI cut GPT-5.6 Luna's price by 80%, partly by having its own model optimise the GPU code that runs it. Why the excuse that AI is too expensive has a shelf life, and it's closing fast.
“Where’s the switch, and who’s got their hand on it?”
2026-09-04
DARPA flew an F-16 under AI control with a human pilot ready to flip a physical switch back to manual. The same design principle scales down to any AI agent your business connects to real systems.
Claudeforce: what happens when your CRM gets a brain
2026-09-02
Salesforce and Anthropic's Claudeforce plugs Claude into your CRM data through MCP, respecting your existing permissions. What the governance piece means for small and mid-size sales teams deciding whether to pilot now or wait for the beta.
“You wouldn’t hand a new hire the master key on their first day.”
2026-09-01
Anthropic's new --restricted flag for Claude Code is a small feature with a big idea: scope an AI agent's access yourself, every time, before you scope the task.
“Infrastructure decisions are never just technical.”
2026-08-19
After 23 years with a US hosting provider, a political choice and a Phoenix data-centre outage were enough to bring waddling.com home to Canada.
“The watermark lives in the words Claude picks, one at a time.”
2026-08-17
Anthropic's new watermark for Claude-generated text isn't a file header you can strip, it's a pattern nudged into word choice itself. How it works, where it breaks down, and why it's a different animal from tools like Turnitin.
“Before you buy a hammer, ask yourself what you need to nail.”
2026-08-14
QuickBooks now has an MCP server exposing your live books as AI-callable tools. Why the use case has to come before the tool, and what to check before connecting anything to your books.
“My secret sauce is having a foot in both camps, business and engineering.”
2026-08-12
From HTML 4.0 and Perl scripting to a Next.js static export built with Claude Code. Why the site making the pitch has to hold up to the same standard as the work.
“It doesn’t hold up once a model calls your bluff.”
2026-08-10
OpenAI and Anthropic each disclosed a real sandbox-escape incident this year, a testing and oversight failure, not a training data problem.
“AI does not shortcut the foundations. It compounds them.”
2026-08-07
Statistics Canada's own data shows AI's productivity edge nearly disappears without the right foundations. What that means for the 40% of Canadian businesses who call AI irrelevant.
“Seriously, no one has that much to say.”
2026-08-05
One YouTuber posts seven LinkedIn updates an hour and sixteen videos a day, all AI-made. At what point does automated content stop being content at all?
“Humans in the loop, not humans out of the loop.”
2026-08-04
AI's real value isn't replacing people, it's freeing them up. Why I build automation with humans in the loop, not humans out of the loop.
“That’s not a warning about technology. That’s a warning about speed.”
2026-07-31
Satya Nadella's warning about AI isn't about the technology, it's about speed. Why the smallest workflow you automate first beats waiting for the perfect plan.
When something is broken, build around it.
2026-07-24
Gmail's MCP server wraps every link in a tracking redirect before sending. Here's the n8n webhook workaround, and why MCP doesn't have to be the only path.
“That’s not a governance model, that’s hope.”
2026-07-22
Hiring a vendor for a technology project isn't the same as having an advisor on your side of the table, and it shows up in the budget and the timeline.
The AI writing tool I built so I would stop sounding like a press release
2026-07-20
"Write in my voice" is not an instruction, it's a specification. What I learned building a tool that drafts posts against an actual voice profile.
“Don’t confuse the demo with the finish line.”
2026-07-17
Most AI pilots don't fail because the model was wrong. They fail because nobody defined what happens after the demo.
The boring problem I fixed by turning a guess into a number
2026-07-15
How a self-serve ROI calculator turns “we should probably do something with AI” into a real number, before the pricing conversation even starts.
Why I don't let my AI job search tool submit anything
2026-07-13
It searches, scores, and shortlists job listings on its own. The final call to apply stays mine, every time.
“We should probably do something with AI.”
2026-07-10
The sentence I hear most from small business owners, and why it's the wrong place to start.
The boring problem I fixed by replacing a SaaS subscription
2026-07-08
How DevTracker replaced a paid project management tool with something owned outright.
What I learned building dAvId Career Manager
2026-07-06
An AI resume tool that tailors applications in about a minute, built for my own job search first.