Forlais Research
FCN Compression
We asked whether large neural assets could take less room with contents still exact. They can, and how much you gain varies more than we assumed.
The question
Less storage with exact contents preserved
The idea was uncompromising on purpose. Not smaller with acceptable loss, but smaller with the contents exact. If a single part came back different, the result did not count.
Research journey
How the work unfolded
FCN Compression achieved lower storage while preserving exact contents. Its retained successful outputs carry complete evidence of restoration, because the claim is worthless without it.
The success carried across several related families of neural asset, and repeated historical outputs stayed consistent within the comparisons we assessed.
Those comparisons showed the opportunity differs materially between related assets. We recorded the variation rather than averaging it away and assuming the same gain everywhere.
Within one specific assessed setting we measured a practical limit, a point beyond which further standalone gains did not continue in the way our early hopes had suggested.
Turning point
What changed the direction
The limit sharpened the work. We recorded where the gain stopped in that setting instead of turning an early hope into a universal claim, and the programme was better for it.
Achievement
What Forlais established
FCN Compression established exact storage reduction with complete assessed restoration across several related families of neural asset, together with an honest account of both the opportunity and the limit.
Significance
Why it matters for AI
FCN Compression is our evidence that valuable neural assets can require less storage while the information inside them stays exact, with variation and limits measured rather than assumed.
Public scope
Forlais Research is the public research surface for Forlais Group. Related official routes include the company site, EvaEsi overview, Project Eve, and selected research programme pages.