I joined Sami Birnbaum on thoughtbot’s Giant Robots Smashing Into Other Giant Robots to talk about building software and checking its work.
Episode link coming when the episode is published.
I’m Ed Woodcock, founder of Bannermedia, where I build software products and work as a fractional CTO.
AI sent me back to Rails.
I genuinely believed AI freed me from the stack. I built in four other stacks in 2026 to prove it. The data dragged me back to Rails.
- Tokens are the new build time. Off the golden path, you pay per token to reinvent
has_many. - Training-data density. The model has seen a million idiomatic Rails apps, so it is better at Rails first time.
- Conventions are a machine affordance. The model predicts where everything lives without reading the repository.
Native mobile, desktop and latency-critical work stay outside, while any mature, convention-rich framework gains some of the same benefit. The framework wars got refereed by an unexpected judge, and the judge charges by the token.
Running a software house where the staff are models.
I don’t pair with AI. I run a software house where the staff are models. Written briefs like tickets, adversarial review of everything, and I read the diffs, never the report.
- “Exit code 0 is not success.” I verify that the expected output exists and meets the brief.
- “Read the diff, never the report.” I inspect the changes because a confident summary can hide omissions.
- “TDD matters more with AI.” A failing test first keeps the human as the oracle while the machine types faster.
Cheap models type; the expensive model’s edge is judgement. I remain responsible for the specification, verification and release.
The sin isn’t AI writing code, it’s nobody reading it.
The sin isn’t AI writing code; it’s nobody reading it. Unreviewed output is inventory, not product.
A working demonstration starts the review. The gate covers failure, maintenance and unexpected use.
Claims that AI has replaced workers need evidence too:
- In a survey of 1,000 US hiring managers, 59% said they emphasised AI when explaining freezes or layoffs. Only 9% reported that AI had fully replaced roles. Resume.org survey, now hosted by ResumeTemplates .
- WIRED reported on 09/02/2026 that New York notices covered 162 employers and nearly 28,300 workers, with zero AI attributions. WIRED’s report .
- A Federal Reserve comparison of late-2025 surveys put adoption at about 18% of firms, 41% of workers and 78% on an employment-weighted basis. Federal Reserve analysis .
Everything I quote here came out of my own AI news pipeline. The tool under discussion did its own homework.
AI eats the grid.
Britain queued an extra Britain. Its data-centre connection queue was about 50 GW across roughly 140 sites, against a 45 GW national peak on 11/02/2026.
A connection queue measures requested capacity, not eventual consumption.
Scroll sideways for the source links.
| Figure | What it means | Source link |
|---|---|---|
| About 50 GW across roughly 140 sites; 45 GW peak | Britain’s data-centre queue compared with peak demand on 11/02/2026. That day’s peak is a reference point, not an all-time record. | The Register, citing NESO and Ofgem |
| 41 GW to 125 GW; at least 80 GW from data centres | Britain’s wider demand-connection queue grew between November 2024 and June 2025. The larger total includes other kinds of demand. | Ofgem , dated comparison |
| £237,500 to £712,500 per MW | Ofgem’s proposed commitment fee on accepting a connection offer. The proposal refunds it at energisation and forfeits it on early exit. | Ofgem’s consultation announcement |
| More than 4 times | Reported UK industrial electricity prices against the US and named Nordic countries, in coverage of OpenAI’s Stargate UK pause. Regulation also mattered. | The Next Web |
| Up to 35 data centres; £10 billion planned | Slough’s cluster faces grid constraints, while Blackstone’s proposed QTS campus near Blyth received outline planning permission. These are separate developments. | Slough reporting , Blyth planning report |
| Ireland: 5% to 23% | Data centres’ share of metered electricity rose between 2015 and 2025. This measures electricity consumed, unlike a connection queue. | Central Statistics Office |
| Denmark: about 60 GW against roughly 7 GW | Requested new demand compared with national peak demand. Energinet describes the pressure on grid expansion and warns that many projects may never materialise. | Energinet’s own explanation |
We design the efficient kit and export the datacentre. Expensive electricity pushes compute towards places where power costs less.
Water is a local question.
Not a global water story, a local one. One big evaporatively cooled site can become a major customer of a small, stressed water system, and the numbers stay secret until someone forces them out.
Withdrawal is water taken from a source; consumption is the part not returned, often because it evaporates.
Scroll sideways for the source links.
| Figure or record | What it means locally | Source link |
|---|---|---|
| London and Slough: drought investigation in 2022 | Thames Water examined data-centre demand and alternatives to drinking water. The report establishes the investigation, without quantifying a resulting shortage. | Thames Water reporting |
| The Dalles: 29% in 2021 | Google’s data centres accounted for this share of the city’s water consumption. Records became public after the city settled its lawsuit against journalists. | Released water records, reported by DCD |
| Google: about 23 billion litres in 2023 | Reported consumption across Google’s data centres. This fleet total cannot identify the pressure on an individual water supply. | Google’s environmental report |
| Newton County: 1 million US gallons daily in 2030 | Appendix A projects this combined demand for Meta’s 2 developments. The document also models lower demand; these are forecasts, not measured use. | County water analysis |
| Mesa: air cooling | Google says its Arizona facility uses air cooling. Local design changes the demand, even where the surrounding region faces water stress. | Google’s Mesa account |
The Newton County study has no baseline for neighbours’ reported well problems, so their cause remains disputed in the reporting on the complaints . The Dalles shows why public records and scrutiny of cooling choices matter.
Family history with an AI that shows its homework.
Family history gave me another reason to insist on evidence. A plausible link can pull an entire branch into the wrong family. I tie each claim to its source, mark inference and preserve contradictions.
The colonel who was also a lawyer.
My great-aunt Gwen went to America as a secretary, sponsored by an army colonel. Later records described her employer as a lawyer in Flint, Michigan. An army lawyers’ journal showed that the colonel was a military lawyer, so both descriptions named the same man.
The brother with his mother’s surname.
Elizabeth grew up with a wealthy, childless uncle while her parents lived nearby with her brothers. A census revealed an older brother whose birth registration used his mother’s surname. He was born before the marriage and stayed with his parents; Elizabeth was born afterwards and lived with her uncle.
The wealthy man who became the wrong father.
A newspaper described Elizabeth as a magistrate’s “niece and adopted daughter”. He entered the tree as her father. The review caught that leap, so I removed the parent link and recorded the relationship as unknown.
The branch resting on a stranger’s guess.
One family-tree website entry supported a branch containing 24 people, 217 documented facts and a New Zealand prime minister. The archive froze the branch until an 1833 baptism register settles the connection.
The AI’s superpower wasn’t imagination. It read everything, connected what I’d never connect, and kept a list of which of its own conclusions not to trust.
Read the family-history guides or take the free family-history course .
What Bannermedia makes.
- Lowdown : live. A daily news digest built by an AI pipeline I run.
- Plotbeacon : beta 2026. It monitors public planning portals and helps residents understand applications and the time available to comment.
- Accountably : in development. A focus tool for working alongside friends without a chat feed demanding attention.
- Lettergraph : in development. A newsletter publishing platform for writers who want to keep control of their audience.
For fractional CTO work, visit wroughtwith.ai . I help teams make technical decisions and review the software those decisions produce.
Get in touch.
Contact me about a product, a technical problem or a source that changes a claim on this page.
Find me on LinkedIn for professional contact.
If a source changes a claim on this page, tell me and I will change the page.