# Сопутствующие статьи по теме Data

The Self-Destruction of the Startup Bible: The More You Know, the Sooner You Fail

The article "The Self-Defeating Nature of Startup Dogma: The More You Know, The Sooner You Fail" argues that popular startup methodologies—such as Lean Startup, Customer Development, and the Business Model Canvas—have not improved startup survival rates over the past 30 years, based on U.S. government data. The core paradox is that once a methodology becomes widely adopted, it loses its competitive advantage as all founders converge on the same strategies, leading to homogeneity and increased failure rates in competitive markets. The author compares this to the Red Queen effect in evolutionary biology, where continuous adaptation is necessary just to maintain position. Despite the intuitive appeal and scientific claims of these frameworks, empirical data shows no improvement in the survival rates of either general U.S. businesses or venture-backed startups. In fact, the success rate for seed-funded startups securing subsequent funding has declined. The article explores three possible explanations: the theories might be fundamentally flawed; they might be too obvious to require formalization; or they might be self-defeating when universally applied. The author calls for a truly scientific approach to entrepreneurship, one that embraces experimentation, paradigm development, and differentiation rather than dogma. The conclusion is that to succeed, founders must often do the opposite of what popular playbooks advise.

marsbit03/23 08:13

The AI Cycle Is Here: Should Web3 Entrepreneurs Pivot to AI?

The article discusses whether Web3 entrepreneurs should pivot to AI amid the growing AI trend, using the metaphor of "raising lobsters" to symbolize the current hype. It notes that while Web3 projects are exploring AI agents for blockchain interactions, identity systems, and payments, many teams are merely chasing narratives rather than building sustainable products. The author cautions against abrupt shifts to AI, highlighting the high barriers, intense competition, and the risk of abandoning hard-earned Web3 expertise. Instead, they suggest a hybrid approach: leveraging Web3’s strengths—such as decentralized data verification, identity management, and micro-payment systems—to address AI’s unresolved challenges like data provenance, agent collaboration, and automated settlements. Key considerations for Web3 teams include assessing their core competencies (e.g., data infrastructure, identity protocols, or application layers), identifying real-world use cases with paying users, and evaluating existing resources (data networks, developer ecosystems) to avoid building in isolation. The goal is to abandon narrative-driven pivots and focus on practical integrations where crypto and AI complement each other.

marsbit03/19 00:33

Using AI for Weather Prediction: Earn $200 a Day While Doing Nothing?

Using AI for Weather Prediction: Can You Really Earn $200 a Day? This article explores how to leverage AI and data analysis to profit from weather prediction markets like Polymarket, focusing on Shanghai’s temperature forecasts. The system relies on Shanghai Pudong Airport (ZSPD) weather station data, sourced via Wunderground, rather than general city forecasts. Key insights include: - Temperature data is reported in whole Fahrenheit values in METAR format, not Celsius, affecting precision. - Historical data shows daily high temperatures most frequently occur between 11:00-13:00, peaking at 12:00 in summer (27.6% of days). Three effective prediction methods were implemented: 1. **Integrated Forecasting**: Combines Weather Company (WC) and ECMWF model data, weighted by weather conditions (e.g., sunny days favor WC). 2. **Real-Time Correction**: Uses morning temperature rise data and historical patterns to extrapolate the daily high, adjusted for cloud cover and wind. A Kalman filter dynamically weights real-time data vs. forecasts. 3. **Temperature Trend Model**: Predicts whether the day will be warmer/cooler than the previous day using pre-dawn data (pressure changes, wind, cloud cover, recent trends). It performs best in winter (clear signals) but poorly in autumn (63.7% accuracy). Two failed methods—Fourier analysis (systematic underestimation) and ERA5 peak-time prediction (insufficient precision)—were discarded. Case studies demonstrate the system identifying mispriced market opportunities, such as recognizing nighttime warming from moist air during rainfall, when public sentiment lagged. Limitations include autumn inaccuracy, lack of real-time pressure data, and unresolved coastal wind effects. Ultimately, the goal isn’t perfect accuracy but leveraging informational edges when odds are favorable.

marsbit03/18 12:18

x402 Wish List

"x402 Wish List" explores the potential of x402 as a micro-payment and data aggregation protocol, based on a Galaxy Research report. It highlights a use case where an AI agent uses x402 to query weather data, book trips, and manage payments to various data sources. The author envisions x402 enabling new business models, such as pay-per-use "Skills Endpoints" for AI agents, replacing fixed fees with usage-based pricing. Other proposed applications include curated data streams for niche crypto news, aggregated ecosystem metrics, a tracker for media prediction performance, and a security/audit incident aggregator. The summary concludes by questioning the economic viability of micro-payments at scale and the legal challenges of data sourcing, noting that x402 V2's revenue-sharing features could help align incentives between data aggregators and original providers.

marsbit03/17 12:24

Bitcoin Rebound Driven by ETFs and Long-Term Holder Accumulation

TheNewsCrypto03/16 12:34

Code Is Getting Cheaper, Licenses Are Getting More Valuable: The Real Moat of Fintech in the AI Era

"The article argues that in the AI era, code is becoming cheaper, but true competitive advantages in Fintech come from hard-to-replicate elements like banking licenses, underwriting data from credit losses, and proprietary risk models trained on real transaction volumes. While traditional software companies face compressed margins as AI lowers both development and distribution costs, Fintech's real moats lie in regulatory compliance, risk management, and data accumulation—areas where AI actually strengthens incumbents. Companies that own the "fin" side (licenses, risk-bearing, proprietary data) will benefit from AI, while those relying solely on "tech" (software interfaces built atop others' infrastructure) will suffer from eroding differentiation. The new arbitrage is no longer about applying software valuations to financial economics, but about owning the foundational elements of finance itself."

marsbit03/16 04:56

Conversation with CoinGecko CEO: Building a Crypto Data Platform from Scratch with $200, Regularly Evaluating Acquisition Intentions

Bobby Ong, co-founder and CEO of CoinGecko, built the cryptocurrency data platform from scratch with just $200 over 12 years, without venture capital. In a recent interview, he shared his journey into crypto in 2013, driven by an economics background and interest in Bitcoin's fixed supply and self-custody model. He remains optimistic about NFTs, viewing them as essential for tokenizing real-world assets, though he believes future crypto winners have yet to emerge. Ong discussed CoinGecko’s origins, including its name choice for being adaptable and fun, and highlighted the 2021 DeFi boom as a key growth driver, as the platform was early to track DeFi tokens. He explained the company’s decision to avoid VC funding, which fostered creativity and disciplined, low-cost marketing like SEO optimization. When comparing CoinGecko to competitor CoinMarketCap, Ong acknowledged CoinMarketCap’s stronger SEO due to earlier market entry but emphasized CoinGecko’s broader token coverage and data accuracy. He also addressed rumors about a potential $500 million sale, stating that while CoinGecko regularly evaluates strategic options, including acquisitions, there is no perfect timing for such decisions—it depends on the team and circumstances.

marsbit03/15 07:15

Your Backtest Is Lying: Why You Must Use Point-in-Time Data

The article warns that backtests using revised historical data are misleading due to look-ahead bias. It illustrates this with a hypothetical trading strategy based on Bitcoin exchange outflows from Binance: entering the market when the 5-day moving average of BTC balance falls below the 14-day average, and exiting on the reverse. Initial backtest results using current data appeared reasonably successful, even comparable to buy-and-hold at times. However, the author argues this is deceptive because metrics like exchange balances are often revised retroactively as new data arrives, meaning a backtest uses information that wasn't available in real-time. Re-running the identical strategy using Point-in-Time (PiT) data—which is immutable and reflects only what was known when the data was first published—yielded significantly worse performance. The PiT-based strategy failed to capture key market upswings as effectively, resulting in meaningfully lower cumulative returns. The key takeaway is that only immutable, Point-in-Time data ensures a backtest accurately replays history without look-ahead bias, and that using revised data will produce lying backtests.

insights.glassnode03/13 14:12

Your Backtest Is Lying: Why You Must Use Point-in-Time Data

This article warns against a common pitfall in backtesting trading strategies: look-ahead bias caused by using revised historical data. It illustrates this with a hypothetical Bitcoin strategy based on exchange outflows from Binance. The strategy is built on the premise that sustained outflows (when the 5-day moving average of BTC balance falls below the 14-day average) are bullish, while inflows signal a sell-off. An initial backtest using standard, revised data shows the strategy performing comparably to a simple buy-and-hold approach. However, the author argues these results are misleading because the data has been updated with information that wasn't available in real-time. This data mutation creates an unfair advantage in the backtest. To demonstrate, the test is rerun using Point-in-Time (PiT) data—an immutable, append-only record that reflects only what was known on any given day. The results are significantly worse, as the PiT-based strategy misses key profitable moves. The key takeaway is that accurate backtesting requires immutable Point-in-Time data to avoid look-ahead bias and replay history honestly.

insights.glassnode03/13 14:12

Pharos Network Expands RealFi Alliance, Joins Forces with Strategic Research and Intelligence Partners to Enhance Institutional Transparency

Pharos Network, a RealFi inclusive financial Layer 1 project, has expanded its RealFi Alliance by adding strategic "Smart Partners" focused on research, data, and institutional infrastructure. The new partners include Dune, Four Pillars, Web3Caff Research, Anchorage Digital, Alchemy, Aquaflux, and Yield Network. These collaborations aim to address information asymmetry and enhance institutional transparency in decentralized markets. Key initiatives include in-depth institutional research, real-time verifiable dashboards for RWA fund flows, enterprise-grade RPC infrastructure, federally regulated banking and custody services, on-chain treasury management, and improved capital efficiency. A core objective is to establish a standardized "RealFi Research Framework" to measure RWA performance, risk, and compliance. According to Pharos CEO Wish Wu, the partnership ensures that assets on Pharos receive the same level of scrutiny and transparency as traditional capital markets. Pharos Network is a modular Layer 1 blockchain designed for RealFi, combining institutional-grade asset circulation with decentralized composability. It is developed by a team with backgrounds from Ant Group and backed by investors like Hack VC and Faction VC.

marsbit03/13 12:26