Independent AI magazine
We put these tools on real tasks, then write down what held up, what quietly fell apart, and how to tell the two apart before you depend on it.
Front page
Most advice starts with a tool and looks for a use. Start with the task instead, and put it through three questions. Two of them are about checking, and that is the point.
Read the articleSections
Each section holds a question worth returning to, not a stream of product news.
Which tasks it genuinely takes off your plate, and which ones it only appears to.
02What a team of three to twenty people gets back in hours, against what the subscriptions cost.
03Designing chains of steps that fail loudly and safely instead of quietly and expensively.
04Plain answers to the questions the vocabulary makes harder than they need to be.
05The failure modes, the displaced costs and the job effects, without the drama.
Latest
Newest first. Every piece states what was tested and what was not.
Wrong price, line never ordered, wrong reference, wrong currency: caught 383 times out of 383. All seventeen wrong approvals of the test sit on one condition, over billing, and one model wrote the discrepancy down before approving it.
Read the articleSelf-checking catches 97.6 % of bad invoice extractions. On contact merges, where nothing can be recomputed, it catches 76.0 %. Cross-model agreement drops from 92.9 % to 62.9 %, and it works worst on the model that makes the fewest mistakes.
Read the articleOn clean invoices the upgrade buys nothing: all six models are at 100 %. Where it does buy something, all three vendors buy the same 10 or 11 correct invoices out of 96, and the price of that identical gain varies by a factor of 6.5. Costs read from the API, never from a price list.
Read the articleThe arithmetic check we published in August catches 59 % of the errors and fires on 22 % of the correct extractions. We put four different guardrails on the same 288 extractions. The one that catches the most is the worst one to deploy, and cross-model agreement never fires on a clean invoice.
Read the articleThe same input, the same model, the same instruction, ten times at temperature 0. A third of the pairs came back with more than one answer, and every single divergence passed a schema check. One model out of three was stable everywhere, which breaks the rule you were about to quote.
Read the articleThree cheap models, 8 meetings, four transcript formats. They find 94.8 percent of the real actions. What they add is the problem: work cancelled during the meeting comes back on one list in five, with somebody's name on it.
Read the articleNo vendor pays to appear, to rank higher, or to be left out. When a tool is named, it is because it was used.
When a piece gets something wrong, the correction goes on the article itself, dated, not quietly edited away.
Anything we did not run ourselves is marked as such in the text. A claim without a test behind it says so.
One email when a new article is published. No tool of the week, no affiliate list, no forwarding of your address to anyone.
