The compression algorithm delivers up to 38% token reduction per context call - benchmarked on real prose content. Total savings compound: fewer calls, precise context assembly, smaller prompts on every step.
+
Neuron delivers a 40% reduction in tokens across your AI interactions. Precise context assembly means smaller prompts on every call. Total savings compound: fewer calls, less compute, lower cost on every step.
- The AI that remembers doesn't have to recompute.
+ 40% fewer tokens. Same work done.
- Every time you open ChatGPT and explain who you are again, that's computation that didn't need to happen. Every time the AI re-establishes your context from scratch, energy is spent re-deriving what it already knew.
+ Neuron delivers a 40% reduction in tokens across your AI interactions. Less computation. Lower cost. A smaller footprint. This isn't a setting you toggle - it's what persistent context does by default.
- Neuron's memory graph persists. The context tax - the repeated re-inference of who you are and what you're working on - doesn't accumulate. Over months of use, that compounds into a meaningful reduction in total computation.
+ Every time you open ChatGPT and explain who you are again, that's computation that didn't need to happen. With Neuron, that context tax doesn't accumulate. Over months of use, 40% fewer tokens compounds into a meaningful reduction in total compute - and a meaningful reduction in what you pay.
-
+
This isn't a green marketing claim. It's a consequence of the design. The same architecture that makes Neuron better for you also makes it lighter on the planet.
+
+
+
Savings calculator
+
If you spend $50/month on AI…
+
+
+ $240
+ saved per year
+
+
Based on 40% token reduction applied to your monthly spend.
- Your memory graph lives in a custom on-device storage layer we built for this purpose. No cloud database servers running 24/7 to store and serve your conversations. No replication across availability zones. Just your device.
+ Your context lives on your device in a purpose-built local storage layer. No cloud database servers running 24/7 to store and serve your conversations. No replication across availability zones. Just your device.
Persistent context = less recomputation
-
Remembered once, retrieved - not re-derived
+
No re-explaining. No wasted tokens.
- The structured memory graph means Neuron retrieves relevant context rather than re-inferring it from long conversation histories. Shorter, more targeted prompts. Less total token processing per useful outcome.
+ Neuron surfaces exactly what's relevant for each conversation - no re-deriving who you are from long histories. Shorter, more targeted prompts. 40% fewer tokens per useful outcome.
Every AI resets when you close the tab.
- I built the one that doesn't - and priced it below the ones that do.
+ I built the one that doesn't.
- Runs on your machine. Remembers everything. Cheaper than ChatGPT on day one.
- No cloud required - and when you want inference, mine costs less than theirs.
+ Runs on your machine. Remembers everything. Priced below ChatGPT on day one.
diff --git a/src/inference.el b/src/inference.el
index 5bb6aed..43f224a 100644
--- a/src/inference.el
+++ b/src/inference.el
@@ -43,7 +43,7 @@ fn inference() -> String {
NeuronThe model - live on day one
-
Built on Qwen3 - open weights, Sonnet-equivalent capability. Neuron wraps it with your full context, your memory graph, and years of accumulated intelligence. The model is the floor. Neuron is the ceiling.
+
Built on Qwen3 - open weights, Sonnet-equivalent capability. Neuron wraps it with your full context, years of accumulated intelligence. The model is the floor. Neuron is the ceiling.
![\"Neuron\"](\"/assets/brand/neuron-wordmark-on-light.png\")
diff --git a/src/efficiency.el b/src/efficiency.el
index 77e80a7..dd21281 100644
--- a/src/efficiency.el
+++ b/src/efficiency.el
@@ -57,10 +57,10 @@ fn efficiency() -> String {