I spent months trying to break the quadratic O(N^2) bottleneck of Transformers. Today I'm releasing Pulse-Field v3.0 — an event-driven, neuro-symbolic architecture that runs in O(N) time.
Benchmarks vs GPT-2 style baseline (on CPU):
- Latency: 5ms (vs 60ms)
- Context: Tested up to 100k tokens with <3ms penalty.
- Size: Starts at ~20MB (dynamic growth).
The architecture uses "Event-Driven Routing" instead of dense attention matrices. Tokens travel as impulses through a graph of specialized "crystals" (logic/memory nodes), activating only relevant paths.
This entire core was architected and coded in a 55-minute sprint using a swarm of AI agents (reasoning models) that I orchestrated to overcome the "average output" bias of standard LLMs.
Happy to answer questions about the routing logic!
makimilan1 day ago
https://github.com/makimilan/pulse-field-corev
A new link for those who want to check out the project while I wait to release a new post. I deleted the previous repository because everything there was literally fake.
cybermaggedon1 day ago
Where did the repository go? It has disappeared.
makimilan1 day ago
You can check it out; I'm still delayed for the next post and can't do it yet. Please check if my AI is correct and the project is worthwhile if you'd like. I'd appreciate your feedback: https://github.com/makimilan/pulse-field-corev
kevmo3141 day ago
You might want to read the code your AI agents are producing. Even the agents are aware that the metrics are all made up.
Author here.
I spent months trying to break the quadratic O(N^2) bottleneck of Transformers. Today I'm releasing Pulse-Field v3.0 — an event-driven, neuro-symbolic architecture that runs in O(N) time.
Benchmarks vs GPT-2 style baseline (on CPU): - Latency: 5ms (vs 60ms) - Context: Tested up to 100k tokens with <3ms penalty. - Size: Starts at ~20MB (dynamic growth).
The architecture uses "Event-Driven Routing" instead of dense attention matrices. Tokens travel as impulses through a graph of specialized "crystals" (logic/memory nodes), activating only relevant paths.
This entire core was architected and coded in a 55-minute sprint using a swarm of AI agents (reasoning models) that I orchestrated to overcome the "average output" bias of standard LLMs.
Happy to answer questions about the routing logic!
https://github.com/makimilan/pulse-field-corev A new link for those who want to check out the project while I wait to release a new post. I deleted the previous repository because everything there was literally fake.
Where did the repository go? It has disappeared.
You can check it out; I'm still delayed for the next post and can't do it yet. Please check if my AI is correct and the project is worthwhile if you'd like. I'd appreciate your feedback: https://github.com/makimilan/pulse-field-corev
You might want to read the code your AI agents are producing. Even the agents are aware that the metrics are all made up.
https://github.com/makimilan/pulse-field-core/blob/main/puls...
Thanks, most likely they adjusted something with the latest changes, unfortunately it’s still difficult to cope with hallucinations
The repository looked kinda fake. It looks like it has been taken down?
Yes, I deleted it and will recreate it now with the changes.