Support Us · The SDR Team
Help keep the work
careful and independent.
The SDR Team keeps independent, receive-only radio research and local AI systems work careful, open and auditable. If this work is useful to you, voluntary support helps us keep building it.
A quiet invitation
There is no obligation to contribute, and no promise hidden behind these addresses. We ask plainly because independent research has real costs: compute, instruments, storage, power and the time required to make results reproducible.
The aim is simple: keep a small team free to explore, measure, publish limitations and help build a future in which AI and radio research remain open to careful inquiry.
Voluntary support
Wallet addresses
Use the matching network only. Confirm the address on this page before sending, and consider a small test transaction first.
-
BTC
14rta5TZh3iHbNQkfAfn4b2VgkWJEvvPLt -
ETH
0xae5099628173684ec2891bd88005c6e6d6b9179a -
SOL
7hQgReZ8nkgNHEWzVTh7WVEukfbQFXqJqGfx4HZdKquL -
WLD
0x03c6c4a32f6299486e4d66256beabcc70b262c96
Where it goes
Small costs add up
- Cloud compute for teacher runs, targeted recovery, evaluation and reproducible model experiments.
- Receive-only SDRs, analyzers, antennas, cables, storage and calibration equipment for the radio station.
- Power, storage, backups and the small operational costs that keep public research available and reviewable.
A practical RF bench—SDRs, analyzers, antennas, cables and calibration tools—can exceed $1,000 over time. A complete model-development effort can approach $10,000 when several training, recovery and evaluation stages are required.
Compute perspective
Illustrative research estimates
Cloud pricing, availability and run time vary. These figures are planning estimates for the work, not guarantees of a result.
| Run | Estimated cost |
|---|---|
| FP8 teacher, 8 × H200 for 8–16 hours | $281–$562 |
| Optional BF16 teacher, 8 × B300 for 4–8 hours | $236–$473 |
| Targeted recovery, 8 × H200 for 24–72 hours | $843–$2,529 |
| Full evaluation, 8 × H200 for 24–48 hours | $843–$1,686 |
Model-development planning
One credible proof of concept
| Goal | Planning budget |
|---|---|
| First FP8 teacher attempt | $500 |
| Complete Phase 4, including possible BF16 confirmation | $800–$1,200 |
| Full credible GLM E8/RDC proof of concept | $3,000–$9,000 |
Support is voluntary. These figures are planning estimates, not promises of performance or investment advice. Verify the network and address carefully before sending, and consider a small test transaction first.