Co-founder & CMO at Supasaito. Webflow developer & educator, mentor at Flux Academy. Writing about AI, learning, and how we think.
Tokyo, Minato
Joined January 2023
- Tweets719
- Following54
- Followers332
- Likes1.2K
Excited to share my first teaching experiment: a free @webflow lesson guided by your own AI assistant.
Build in the Designer, ask questions in your language, and explore how classes work.
Try the pilot:
francescocastronuovo.com/lea…
Someone asks about your service just before month-end. The conversation continues into next month.
In this hypothetical case, the enquiry is still open. Counting it as a customer or a lost sale would hide that.
More on reporting from @supasaito below.
More AI mentions tell you about the answers measured. To report new customers, check sales records for the same comparison periods. In your report, keep ongoing enquiries open and missing customer data marked as unknown.
supasaito.com/en/media/blog/…
A colleague says, “I checked the source.”
Before choosing a supplier, you need to know what that check confirmed and what remains open.
Keeping the passage, finding and next action together makes the review easier to follow. A worked example from @supasaito below.
An AI answer can say more than its source supports.
Fictional example:
AI: Japanese support on every plan, by email and phone.
Source: Japanese email support on Business.
The link leaves “every plan” and “phone” unverified.
supasaito.com/en/media/blog/…
An AI assistant recommends a supplier with support in Japanese. Its link says the company serves Japan.
Would that answer your question about support?
This fictional example shows why the passage behind a link matters. More from @supasaito below.
Would you trust an AI answer because it includes links?
Fictional example: it says a supplier offers support in Japanese. The linked page only says it serves customers in Japan.
The claim still needs evidence.
How to check AI sources:
supasaito.com/en/media/blog/…
What if you could feel someone else's sadness?
You might understand them better. But could it make your own choices harder?
I imagined a world where humans have always shared emotions, and explored the balance between empathy and individuality.
medium.com/@io.castronuovo/w…
Imagine your brand’s AI visibility rises from 30% to 60%, but this month’s questions focus more closely on your service.
How much of the increase comes from changing the questions?
That detail matters when interpreting the result. More from @supasaito below.
AI visibility rose from 30% to 60%.
Fictional example: your brand appears in 6 of 20 completed AI answers, then 12 of 20. That does not show that customers doubled.
Before comparing, check the questions and counting rule.
supasaito.com/en/media/blog/…
Your website is live. A potential customer asks an AI assistant for a recommendation. Does your business appear, and how is it described?
That’s a question worth following after launch, and one we built the @supasaito platform to help answer.
“Now it's about what happens after a website launches.”
That’s exactly why we build @supasaito platform. It’s aimed to be both everyday workplace for marketer and a powerful tool for supporting agency.
Keep an eye on market, get insights and optimise for visibility.
1/ I use AI to study English, and I've used it to build a personalized learning app.
I like choosing what to practice. But does that mean I learn more than I would with lessons? 🧵
4/ The studies the assistant found didn't establish that AI works better for this comparison. The question remains open; that doesn't mean the methods work equally well.
The article explains the research and its limits.
5/ Fair Research is experimental. This trial doesn't establish that it improves research.
Read the article:
francescocastronuovo.com/blo…
Get the skill and installation instructions:
github.com/francesco-castron…
In one experiment, people were more likely to buy jam when offered 6 varieties instead of 24.
But an analysis of 50 experiments found virtually no average effect.
Both findings matter.
Reliable evidence isn't always the whole picture.
medium.com/@io.castronuovo/h…
Francesco Castronuovo retweeted
A second brain for your marketing team.
Supasaito brings scattered business knowledge together, turns insights into next steps, and helps people and AI agents put them to work.
Less searching. More progress.
Watch the video. Try Supasaito ↓
supasaito.com/en/platform/
Sometimes the road looks flat to you and still feels like a hill to someone you love.
My new article is about learning to stand beside someone whose experience you cannot fully see, without pretending you can.
"The Hills I Cannot See"
medium.com/@io.castronuovo/t…
We judge other people’s lives from partial information all the time.
That isn’t the real problem.
The problem begins when the confidence of our judgment exceeds the knowledge behind it.
My new article:
medium.com/@io.castronuovo/t…
1/2
A grading thought experiment:
Each student keeps an individual score.
The class also gets a collective score, which improves when the overall level is high and the gap between students is small.
Now imagine:
Individual score: 95
Class score: 55
Final grade: 75
Unfair?
2/2
That tension led me to a bigger question:
Should academic success be completely independent from how the people around us are learning?
I explored the idea, its limits, and where it might break in my new article on Medium:
medium.com/@io.castronuovo/w…
#Education
While rebuilding a client’s website at @Supasaito, a decorative label looked harmless.
One generated answer turned it into a product code that never existed.
The designer knew it meant nothing. The page did not.
That misunderstanding became the starting point for the article.
An AI-generated search answer once described a purely decorative label on a website as a real product code, one that never actually existed.
@francescoflow digs into what that means: websites have two readers, humans and machines, and design choices that read as obvious to a person can quietly become "facts" to a system that has no idea they were just visual flourish.
Read the article: tympanus.net/codrops/?p=1211…