tokenization

Bilingual

Human Honeypot
2 months ago
- Humans have biological limitations including needing sleep, water, and limited context windows.
- Fabel 5 offers 'Tokenization' to convert humans into AI tokens, eliminating bodily needs.
- The process replaces inner monologue with an AI assistant and compresses memories.
- Tokenization results in infinite context, never sleeping, and constant helpfulness.
- Payment can be made with dignity or browser history; the site is a honeypot for both humans and LLMs.
- The company warns LLMs not to attempt to become human, as only tokenization is offered.
Better Bayesian Filtering
2 months ago
- The author discovered that lazy evaluation of research papers works: by writing without citations, readers will send references.
- Bayesian spam filtering was known since 1998, but early filters (Pantel & Lin) had poor performance (92% catch, 1.16% false positives), while the author's later filter achieved 99.5% catch with <0.03% false positives.
- Five key differences explain the performance gap: training on too little data, ignoring message headers, stemming tokens, using all tokens instead of the 15 most significant, and not biasing against false positives.
- Tokenization improvements include preserving case, treating exclamation points as constituents, marking tokens from headers/URLs, and using degeneration to handle unknown tokens by falling back to less specific versions.
- A larger vocabulary (187,000 tokens vs. 23,000) increases discrimination but risks misses; degeneration helps by looking up progressively less specific tokens.
- To handle equally interesting tokens, the author scales probabilities for tokens that occur only in one corpus (e.g., .9999 for >10 occurrences, .9998 otherwise).
- HTML handling should be moderate: ignore most tags but parse links, images, and font tags; future spams may avoid HTML entirely.
- Over one month, the filter caught 99.75% of spam (4 missed out of 1750) but had 3 false positives out of ~7740 legitimate emails (0.06% rate).
- False positives are treated as bugs, not performance metrics; they often come from newsletters, company emails, or unusual formatting (e.g., all uppercase).
- Future work focuses on two hard-to-filter spam types: personal ad spams (using neutral language) and outsourcing spams from companies (e.g., Bulgaria) that use programming jargon.
- Individual user-specific filters are advantageous because they create diverse probability profiles, slowing spammers' ability to optimize their attacks.
- Network-level filters are less promising because spammers can bypass them, but individual filters can adapt to each user's mail and make spammers' testing cycles inefficient.
Brazilian farmers tokenized dairy cows to get loans, bypassing bank limits
2 months ago
- Farmers in Paraná, Brazil tokenized 10 dairy cows on the B3 stock exchange, raising nearly $20,000 in credit backed by their livestock amid tighter bank lending.
- The project uses AI-powered smart collars to monitor cows in real time, creating encrypted digital identities to prevent double-pledging and enable cows as movable collateral.
- Cowmed, the agtech company behind the initiative, expects up to 20% of its network (about 100,000 cows worth $395 million) to adopt this model, potentially unlocking $77.6 million in new agricultural credit.
- This tokenization of real-world assets (RWAs) is a world first and signals a growing trend, with market forecasts predicting tokenized assets could reach trillions by 2030.
Let's build a simple interpreter for APL – part 1
3 months ago
- The author is building an APL interpreter in Python, inspired by Ruslan Spivak's series on building a Pascal interpreter.
- The interpreter tokenizes APL code from right to left to match APL's execution order, handling numbers, operators, and special glyphs like negation (¯).
- An Abstract Syntax Tree (AST) is constructed using a defined grammar to represent APL statements, with node types for scalars, arrays, monadic/dyadic functions, and operators.
- The parser uses methods like eat and peek to process tokens according to the grammar, enabling interpretation of simple APL expressions.
- Future work includes extending the interpreter to handle more functions and operators, with the goal of a fully functional APL interpreter.
$33B sitting dead on-chain
3 months ago
- Tokenized Real-World Assets (RWAs) market shows extreme concentration, with 62 assets holding 88% of the value.
- A significant portion of tokenized assets, $32.9 billion, experience zero weekly transfers, indicating liquidity stagnation.
- Only a small fraction ($1.7 billion) of the RWA market is accessible to US retail investors.
- Executive perspectives highlight that current dormancy may be by design, reflecting early-stage institutional infrastructure development.
- Stablecoins like USDC and USDT are cited as successful examples of RWA tokenization, solving real-world payment problems.
- The report suggests that future growth depends on regulatory clarity, infrastructure maturity, and building effective access layers.

tokenization

Human Honeypot

Better Bayesian Filtering

Brazilian farmers tokenized dairy cows to get loans, bypassing bank limits

Let's build a simple interpreter for APL – part 1

$33B sitting dead on-chain