by Rafael Yashiki Capua, COO & Co-Founder of Jump
Welcome. AI has left a lot of knowledge looking standard, repeated from summary to summary. Here I bring to the surface what comes from real experience and checked sources, in editions built for making decisions.
What you take from each edition
Jump · ISO/IEC 42001 certifiedthe international standard for AI management
01editions
area · AI risk and security
How agentic AI systems get attacked, what fails in practice and what actually holds. 2 editions published · 1 in preparation
what the cases really were
was the price of the expired domain still on Salesforce's allowlist (approximate)
of Cloudflare's own were sitting inside the stolen support cases, pasted there by customers
02method
holds for every topic, in every edition
Vendor documentation, the affected company's report, published research, official data. Every citation carries a date and a link, and the sources close each edition.
A measured fact, a vendor estimate and a lab demonstration weigh differently in a decision. Each one is labelled for what it is.
When the repeated version is wrong, the edition shows the correction and its source. Edition 02 of agentic systems opens by correcting edition 01.
What is documented, what is research and what is my own recommendation stay separate, and each edition says what I could not verify.
03who writes
from executive to rider, from rider to executive
I am Rafael Yashiki Capua, COO and Co-Founder of Jump, a Brazilian data and artificial intelligence company. I run operations and sit at the table with executives at the moment AI and data leave the hype behind and enter company strategy. These editions come from that table.
Outside the company, I race in Brazil's national motorcycle road racing championship. In the pit box, decisions about the bike come from the telemetry, however good the lap felt. I bring that discipline to the company and to every edition.
At the company and on track, you have to run close to the limit without crossing it, align the team around the same data, and make mistakes in practice, where they still come cheap.
This space is personal. Jump comes in at the end of each edition, kept apart from the analysis, with the limits of what it delivers stated out loud.
Company and track demand the same thing from whoever decides.
04ecosystem
Jump solutions, organised by the problem they solve
what AI and data the company uses, and how it proves it is under control
where data runs, what it costs and how to migrate without losing what it says
MigrateMindMigrates legacy analytics workloads with real execution, source-to-target validation and an audit trail.migratemind.jump.tec.brexplore 
Opt3Visibility, optimisation and governance of cloud cost across AWS, Azure, GCP and Databricks.opt3.jump.tec.brexplore 
NautilusIQjump.tec.br/nautilusiqexplore what the team knows and how it stops depending on a single person
how projects and day-to-day operations move, from start to delivery
05contact
The shortest path is a message on LinkedIn. If you found a mistake in an edition, tell me too, because the correction goes into the text with its source.
Rafael Yashiki Capuapersonal profiles
