4,600 people sorted human text from machine text at 50 to 52%
Polish is not a detection cue, because detection barely happens. The rougher cut of my ad did win on hook rate, and I had explained why it won incorrectly.
Mo RezaAli
I'm a Growth Marketer at Yakkyo S.p.A., and I work from Messina, Sicily, Italy.
Growth marketing
The expensive growth decisions are the ones made on numbers nobody checked
I work on growth at Yakkyo S.p.A.: acquisition, lifecycle, and the automation under both. The calls that cost real money are the ordinary ones, made on a number that looked settled. A checkout test needs 8,087 mobile sessions a day for a week to detect a 20 percent lift; on 641 it sees only an 80 percent one, so a flat result there is a guess wearing a chart. An agent chain at 85 percent a step is a 32 percent agent across seven, which is why it demos beautifully and fails in production. A model at half the list price does not cost half if it spends three times the tokens.
The most expensive version of that mistake I have run was my own. I built a pipeline that checked its own output for evidence and could not tell a sourced number from one that merely looked sourced, so it enforced the appearance and published 842 articles nobody read.
What a quality gate cannot see, and how to build one that can
Writing
Latest writing
An always-on laptop draws 3 W to 7 W, which is £7 to £16 of electricity a year
The guess I inherited was 10 W to 30 W. The certified data puts it at 3 W to 7 W: £7 to £16 a year, four or five dollars of model spend, and $0 in licences at this scale.
Automate up to the decision, and never the decision itself
X’s ranker exposes nineteen prediction heads. None is an impression, four are negative, and net-negative posts leave through a different branch. Three automations I killed, eleven I kept.
Chain of thought is worth 0.7 points outside math and logic
A meta-analysis over more than 100 papers: 12.3 points on math, 0.7 on everything else, 56.8 against 56.1. What that leaves of three techniques people still sell.
Google’s checkout form guidance ships an autofill token that does not exist
The string address-line-1 appears five times on web.dev, four of them inside an autocomplete attribute. Also the single-column study that excluded mobile, and why device conversion splits are partly artifact.
K3 lists at half of GPT-5.6 Sol, and the saving is either 9.6% or 30.6%
Artificial Analysis publishes both figures, for the same metric and the same two models, on two of its own pages on the same day. Both are in the table. Neither is endorsed here.
Rules
How this site works
- Every article lists its sources: publisher, title, date, and a link. Editorial policy
- Every source URL is checked against the live web when the site builds. A dead link fails the build. Colophon
- Corrections are logged in public, with the date and what changed. Corrections
Credentials
Things you can check yourself
- Yakkyo S.p.A. Growth Marketer at Yakkyo S.p.A. The company is listed on Euronext Growth Milan under ISIN IT0005573966. yakkyo.com Euronext Borsa Italiana
- Jumbula Marketing Specialist at Jumbula, September 2019 to March 2026. jumbula.com
- Università degli Studi di Messina Undergraduate study in computer science, in progress, at Università degli Studi di Messina. unime.it
- Coursera Online non-credit course certificates, with the verification link for each one. Certifications
These links go to the organisations themselves and to the exchange listing. They confirm that those organisations exist and that Yakkyo S.p.A. is a listed company. They do not prove that I work there or that I am enrolled, and no public registry publishes either fact. If you need better proof than my word, email me: contact@therezaali.com