13 Mar 2026
Good afternoon from sunny Manchester.
This week we're at Tes SEND Show North 2026 and have had a fantastic time talking to exhibitors and attendees and listening to a range of fascinating talks. We thought the “silent disco” approach was really clever, as it allows multiple talks to run simultaneously in close proximity without dominating the rest of the event. Though it did somewhat complicate the speakers' ability to take questions, as everyone was wearing headphones!
In other news, we’ve been exploring ways to anonymise data, using machine learning.
Why is this necessary/good? As we are using AI coding agents to fast-track prototype development, we need a reliable and efficient automated method for anonymising input data. This is essential to ensuring our prototypes are safe to use without requiring manual anonymisation, which is very labour-intensive and tiresome.
We’ve experimented with a number of existing data redaction approaches, such as Textract and Comprehend. Generally, we have been encouraged by their overall performance, though we have identified a number of areas for improvement, especially in identifying proprietary terminology beyond the scope of more common personal data types, such as names and addresses.
To mitigate, we intend to combine automatic anonymisation with manual review, allowing users to review and amend the anonymised data before proceeding. This will ensure we can leverage the approaches in their current form while we work to improve them.
We wish you a fantastic weekend.
