Fatfort
fat-fort is an early-stage effort in adversarial machine learning and computer-vision robustness — auditing where models break, and how to certify that they don’t. Built in the open, honest about the stage: no invented clients, no borrowed logos, no certifications that don’t exist yet.
The mark is a 2×2 grid whose corners read TT · TF · FT · FF. fat-fort is the FF corner (in red); its sibling tutorsfirst is the TF corner — two cells of the same matrix.
What it explores
- Adversarial examples
- tiny, deliberate perturbations that flip a model’s prediction while looking unchanged to a human.
- Certified robustness
- provable guarantees over a bounded neighbourhood of an input — not merely “we couldn’t break it”.
- Honest evaluation
- robustness claims are easy to overstate; the real work is measuring them without fooling yourself.
Pointers
- fatfort.com — the project’s home.
- tutorsfirst — sister project (tutorsfirst.com.au), an Australian tutoring marketplace where every tutor is credential-verified.
- abaj.ai — main site, writing, and past work.
Reach the fort at hello@fatfort.com.
Backlinks (3)
1. λambda /roam/lambda/
λambda — Learner-Adaptive, Marks-Bound Drilling Agent — is a study tool that keeps a working image of your mind: what you hold, where you stall, and which exam marks each gap is costing you. It probes before it teaches, teaches only what you missed, and drills until the fix survives a variant. The protocol is open source at lambda-agent; the hosted study lounge is being built at lambda.fatfort.com.
■ your mind, as λ understands it ■ what the Language Model knows that is relevant to you ■ everything else the Language Model knows
2. TutorsFirst /roam/tutorsfirst/
tutorsfirst is an Australian tutoring marketplace — find a trusted, verified tutor by subject, level and suburb. Browsing is free, tutors list free and keep 100% of their lesson fees, and students pay one small flat connection fee to swap contact details. No subscriptions, no commission.