[ANN][MAINNET] NEB: mining trains a neural network, starts Nov 18, 14:00 UTC

NEB mainnet starts on Wednesday, November 18, 2026, at 14:00 UTC.

After two testnets, the chain starts from block 0 and trains its two networks from the first block: KEP1 on Kepler transit signals, RET2 on retinal scans.

10 % of your testnet coins carry over. The first block pays each address 10 % of what it holds on the two testnets. For the second testnet the balances are read at block 12900, expected on Monday, November 16, around 15:00 UTC. Coins mined after that block are not carried over. There is nothing to claim: the same wallet file spends them, 100 blocks after the start.

Release 1.0.0 comes out on Monday. Node operators, pools and miners install it before the start, with the new neb-datasets.bin (222 MB) and a new data directory. A testnet node does not join the mainnet.

The genesis block holds the hash of a block of an external public chain mined after this post, and its time is the start time. No block can be mined before Wednesday 14:00 UTC.

60 second blocks, 20 NEB per block, 5 % of it to the project, coins spendable after 100 blocks.

The first post gets the downloads and the genesis id on Monday.

Results so far on testnet 2.

The Kepler model of the chain, trained by the miners on NASA-labelled Kepler transit signals, was the last filter of a search through about 2,000 faint stars of NASA's Kepler Bonus data (406,548 stars, published in 2023). It flagged one exoplanet transit candidate, KIC 6691030. An exoplanet astronomer reviewed it: the signal is real, but its shape and its depth by season point more to a binary or triple star system than to a planet. It is not a confirmed planet and nothing about it is validated. Code and data: https://github.com/SilasEigen/neb/releases/tag/science-kbonus-2

Since block 8800 both networks are convolutional: KEP1 on NASA Kepler transit signals (26,000 weights) and RET2 on retinal OCT scans at 64 x 64 (22,280 weights, research only). In simulation they reach about 91 % and 81 to 83 %. On the live chain, measured on November 14 at block 9632: 87.0 % and 67.8 %, still training.

Site and explorer: https://neuralenginebyte.org
GitHub: https://github.com/SilasEigen/neb
Discord: https://discord.gg/B6dTbScw8
X: https://x.com/silas_eigen
Article: https://medium.com/@neuralenginebyte/every-hash-is-a-training-step-the-engineering-behind-neb-6f0a97714ff8
ResearchHub: https://www.researchhub.com/author/8734734
Field kit fund: https://neuralenginebyte.org/#/fund