SEMTOK

Cut your AI costs by 60%. Fit 2.5x more data in every prompt.

---

The Problem

Every token costs money. JSON wastes half of them on {}"":, noise.

JSON:    {"users":[{"id":1,"name":"Alice"},{"id":2,"name":"Bob"}]}
SEMTOK:  #id|name
         1|Alice
         2|Bob

32 tokens → 13 tokens. 59% saved.

---

Who Benefits

Stakeholder	Impact
Developers	Larger context windows, richer AI interactions
Companies	60% reduction in API costs at scale
AI Agents	Faster communication, more efficient swarms

---

Results

Data Type	Token Savings
Tabular	64%
Nested	46%
APIs	45%
Overall	59.5%

Verified across 41 benchmarks. No cherry-picking.

---

Usage

const { SEMTOKFinal } = require('./semtok-final-v7.js');
const semtok = new SEMTOKFinal();

const compact = semtok.encode(yourData);
const stats = semtok.compare(yourData);
// → { savings: "59.5%" }

---

Format

JSON	SEMTOK
null	~
true	+
false	-
"string"	string
{"a":1}	a:1
[1,2,3]	1,2,3

Full spec: SEMTOK_SPECIFICATION.md

---

Research

Built with scientific rigor:

• 9 parallel AI research agents
• 7 iterations of refinement
• 41 comprehensive benchmarks

Full paper: SCIENTIFIC_PAPER.md

---

Author

David Celekli — Principal Investigator

---

License

MIT — Use it anywhere, for anything.