After a Three-Year Hiatus, Peking University Alumna Lilian Weng's Latest Long-Form Article Goes Viral

marsbitPublicado a 2026-06-26Actualizado a 2026-06-26

Resumen

After a three-year hiatus, OpenAI alum and renowned AI researcher Lilian Weng has published a comprehensive blog post analyzing the "Scaling Laws" that have guided multi-billion dollar investments in large language models. Her post, "Scaling Laws, Carefully," reveals these foundational laws to be more fragile than widely assumed. She dissects a key discrepancy between OpenAI's 2020 conclusion (favoring larger models) and DeepMind's 2022 Chinchilla finding (advocating balanced growth of model size and data), tracing it to differences in parameter counting and experimental scale. Furthermore, Weng highlights a critical methodological bug in the influential Chinchilla paper, discovered in 2024, where a loss function averaging error caused premature optimizer termination. The analysis also underscores a fundamental limitation: the laws assume infinite, unique training data, a premise collapsing as high-quality text data is exhausted by 2026-2028. This scarcity is driving the industry toward alternatives like reinforcement learning, test-time computation, and synthetic data. Weng concludes that the future of AI depends not just on more compute, but on rigorously understanding and refining these scaling principles.

Lilian Weng has finally published a blog post after a three-year hiatus.

Just moments ago, a long-form article by former OpenAI Vice President Lilian Weng, delayed for over three years, went viral.

In this blog post titled "Scaling Laws, Carefully," she deconstructs Scaling Laws from start to finish—

The law on which the AI industry has bet tens of billions of dollars is far more fragile than anyone imagined.

One-Minute Summary: What This 10,000-Word Article Says

A single formula has governed the entire industry for five years. Scaling Laws state that "increasing model size, feeding more data, and scaling up compute will improve performance at a fixed rate." It turned AI from alchemy into a calculable business, indirectly directing the flow of hundreds of billions of dollars.

OpenAI and DeepMind gave opposite answers. For the same question of "how to allocate compute budget," OpenAI in 2020 said models should scale faster than data, while DeepMind in 2022 said both should scale proportionally. It was later discovered that the root of the divergence was a difference in parameter accounting plus insufficient experimental scale.

Even the winner's formula has bugs. DeepMind's optimal allocation ratio, copied industry-wide for two years, was found in 2024 during a line-by-line reproduction: the loss function used mean instead of sum, causing the optimizer to stop prematurely, and the output parameters were not the true optimum.

Applying patterns from small models to predict large models requires great caution. This curve was fitted on relatively small models; when extrapolated to trillion-parameter scale, a rounding difference can cause conclusions to diverge significantly. The blog includes an interactive simulator—drag the sliders to see it with your own eyes.

There's an even more fundamental issue: data is running out. The formula assumes infinite data supply, but high-quality text is finite. This is why the entire industry is collectively shifting towards reinforcement learning, test-time computation, and synthetic data.

One Straight Line, Hundreds of Billions of Dollars

As is well known, the core of Scaling Laws can be simply summarized in one sentence—

The larger the model, the more data, the more compute, the better the performance. And this "better" is not random; it follows precise mathematical laws.

Plotting training loss on a log-log scale shows it decreases in a straight line as model parameter count N, data volume D, and compute C increase.

Written as a formula: L(x) = E + A/x^α, where x can be N, D, or C; E is the theoretical optimal loss (the entropy of the data itself); A and α are fitted constants.

Training a model with N parameters on D tokens requires total compute C ≈ 6ND—2ND for forward pass, 4ND for backward pass.

This straight line means performance gains are predictable.

Run a few small models first, fit the straight line, extrapolate to the right, and you can estimate the performance of a large model. No need to actually spend hundreds of millions training the large model to know if it works.

Before this, deep learning was often ridiculed as "alchemy"—knowing what works, but not why.

In 2020, OpenAI's Kaplan published this power law, pulling alchemy into the realm of "predictability" for the first time.

This is the confidence behind all large model companies' massive investments.

But on the most critical advice the formula gives—how to allocate compute budget between model and data—OpenAI and DeepMind gave opposite answers.

The Same Problem

OpenAI and DeepMind Produced Opposite Answers

The conclusion reached by OpenAI's Kaplan team in 2020 was: optimal model size N_opt ∝ C^0.73.

Translated: if compute increases 10x, allocate 5.5x to model and 1.8x to data—models should scale much faster than data.

This directly guided GPT-3's training plan.

A 175-billion parameter model was fed only 300 billion tokens (a token is the smallest unit of text processed by a model, roughly 1-2 tokens per word).

By later standards, this was severely undertrained.

In 2022, DeepMind's Chinchilla team reached the opposite conclusion: N_opt ∝ C^0.50, models and data should scale proportionally.

Engineers later distilled it into an oft-quoted number: optimal token-to-parameter ratio is roughly 20:1.

Then DeepMind staged a head-to-head match.

Their Gopher: 280B parameters with 300B tokens. Chinchilla: 70B parameters with 1.4T tokens. Both models used the same compute.

Chinchilla dominated completely.

A smaller, well-fed model beat a larger, starved opponent.

Industry consensus flipped: from "scale up models" to "most models are undertrained."

0.73 vs 0.50, opposite answers to the same problem, would have you allocate your compute budget in two completely different directions.

The Reason Turned Out to be an "Accounting Problem"

In 2024, two researchers published a reconciliation paper in the top-tier machine learning journal TMLR, tracing this divergence to its root.

The conclusion is laughable.

First reason: they counted parameters differently.

Models have a type of parameter layer called embedding, responsible for converting text into numerical vectors the model understands. In small models, this layer constitutes a large proportion of total parameters—maybe one-third for models with tens of millions of parameters.

Kaplan excluded embedding when counting parameters; Chinchilla included it.

This single difference in parameter accounting was enough to distort the final fitted power-law exponent.

They gave a concise correction formula: N = N_\E + ω·N_\E^(1/3), where N_\E is parameter count without embedding, ω is a constant. For small models, the second term is large, embedding influence significant; as models grow larger, the second term approaches zero, and the two counting methods converge.

Second reason: Kaplan's experimental scale was too small.

Kaplan's largest tested model was only 1.5B parameters, while Chinchilla's experiments scanned up to over 16B. In log-log coordinates, tiny fitting deviations are magnified dramatically during extrapolation.

Using unified parameter accounting, they rederived Chinchilla's formula and discovered a key pattern—

The power-law exponent changes with compute scale. Within Kaplan's small-scale experimental range, the exponent was indeed close to 0.73; but as scale increased, the exponent converged to 0.50.

Kaplan wasn't "wrong"; he was correct within his experimental scope.

But he extrapolated a locally valid rule into a global conclusion.

An accounting issue of how to count parameters, plus insufficient experimental scale, led two top teams to give opposite resource allocation advice.

The entire industry adjusted training recipes based on this conclusion for two years.

Even the Winner Has Bugs

Kaplan was corrected by Chinchilla—that's the standard narrative everyone knows.

But Weng took a step further—Chinchilla's own methodology also has problems.

The Chinchilla paper used three independent methods to cross-validate its conclusion:

Method 1: Vary data volume with fixed model size

Method 2: Plot iso-compute curves (IsoFLOP profiles)

Method 3: Directly fit parameters to the loss formula L(N,D) = E + A/N^α + B/D^β

Three paths pointing to the same conclusion, seeming very robust.

Method 3's mathematical derivation is especially elegant: Optimizing L(N,D) under constraint C ≈ 6ND yields a closed-form solution N_opt ∝ (C/6)^(β/(α+β)). When α ≈ β, the exponent is approximately 0.5, meaning models and data scale proportionally. That's the mathematical origin of 0.50.

In 2024, a team from AI research institute Epoch AI manually extracted raw data points from Chinchilla's paper charts and reran the fitting for Method 3.

Two bugs, each more astonishing than the last.

Bug 1: The loss function used mean instead of sum.

When fitting these five parameters, Chinchilla needed to minimize the gap between predicted loss and actual loss.

The complete optimization objective: min Σ Huber_δ(log L̂(Nᵢ,Dᵢ) − log Lᵢ), where Huber Loss is an outlier-insensitive loss function (δ = 10⁻³), paired with the L-BFGS-B optimizer to search for the optimum.

The problem lies in a detail: they took the mean (average) of Huber Loss per sample, not the sum. Averaging over a few hundred samples compressed the loss value to an extremely small magnitude.

The L-BFGS-B optimizer has a built-in convergence criterion. It stops automatically when the loss value is sufficiently small. Seeing such a tiny value, it mistakenly assumed convergence and stopped.

The optimizer never finished running. The output parameters were not the true optimum.

Bug 2: Key parameters were only kept to two decimal places.

In the Chinchilla paper, two core exponents controlling the power-law shape were retained only to two decimal places.

It seems like harmless rounding.

But when deriving other constants from these rough numbers, the error was amplified exponentially. The final confidence intervals were unreasonably narrow—narrow enough to require precision achievable only after over 600,000 experiments, while they actually ran fewer than 500.

A formula revered industry-wide for two years concealed a bug where the loss function didn't finish running.

Weng's blog also includes an interactive simulator with three sliders controlling loss precision, loss noise, and fitting range.

Each adjustment changes the fitted Scaling Law.

OpenAI's conclusion had local bias; DeepMind's conclusion had methodological flaws. In the AI industry's most important academic debate, both sides had cracks.

Data Is Running Out

The first three sections discussed problems with fitting methods—how to count parameters, how to calculate loss, how many decimal places to keep.

But even if all these problems were fixed, classic Scaling Laws have a more fundamental vulnerability—

They assume each training data point is unique, not repeated, not trained over multiple epochs, presuming you have infinite data.

Reality is, high-quality text data is projected to be exhausted by major labs between 2026 and 2028.

Repeated data training is inevitable; the premise of the classic formula is collapsing.

A large-scale experiment in 2023 trained about 400 models, from tens of millions to 9 billion parameters, with up to 1500 epochs of repeated training.

The core idea is to introduce the concept of "effective data volume" to replace actual data volume—

If you have U unique data points repeated R times, the effective data volume is not U×R, but converted via the exponential decay curve D_eff = U·(1 - e^(-R)). The first repetition still learns much new information; by the fifth, tenth repetition, marginal learning gains approach zero.

They also found a counterintuitive conclusion: excess parameters "depreciate" faster than repeated data. Meaning, with limited budget, running more training epochs is more cost-effective than enlarging the model.

A new paper in May 2026 took a different approach.

They didn't convert to effective data volume; instead, they directly added an explicit overfitting penalty term to the classic loss formula—the more times a model sees the same data, the greater the penalty, and this penalty is tied to model size.

Their complete formula looks like this:

That last red penalty term is key.

R is repetition count, N/U is the ratio of model parameters to unique data volume (how "excessive" the model is relative to data), P, δ, κ are fitted from experiments. More repetitions, larger models, heavier penalty.

The paper's core finding: Large models are more sensitive to data repetition. Training the same data for 10 epochs, a 500M parameter model might still hold up, but a 5B parameter model's performance would degrade much more severely.

Another directly useful engineering finding: Increasing weight decay can significantly alleviate overfitting from repeated training.

This is also why from 2025 to 2026, the industry's attention collectively shifted to three paths to bypass the data wall—

Reinforcement learning: DeepSeek R1, OpenAI o-series, letting models self-play on verifiable tasks like math and programming to generate training signals.

Test-time computation: Not increasing training cost, letting models "think" a few more steps when answering questions to exchange for better performance.

Synthetic data: Using existing strong models to generate new data to train the next generation.

The subtext of all three paths is the same: The pure "scale-up" power law is no longer sufficient.

From Peking University to OpenAI to Her Own Company

Lilian Weng, undergraduate at Peking University, PhD from Indiana University Bloomington.

Interestingly, her PhD focus wasn't deep learning, but network science and complex systems, studying how information spreads in social networks.

After graduation, she first worked in data science at Dropbox, then at fintech company Affirm, before joining OpenAI in 2018.

At OpenAI, Weng's first project was robotics. She was a core contributor to Dactyl, the robotic hand that learned to solve a Rubik's Cube in two years.

Later, she moved to build the applied research team, and after GPT-4's release, was tasked with forming the Safety Systems team, which grew to over 80 scientists, engineers, and policy experts by the time she left.

In August 2024, her title was elevated to VP of Research and Safety; three months later, she announced her departure.

In 2017, soon after starting deep learning, Weng opened a personal blog called Lil'Log, initially just to organize her study notes.

She once said, "Explaining a concept clearly is the best way to test whether you truly understand it."

Nine years later, she's still writing—reinforcement learning, diffusion models, large model agents, each article starting from foundational principles, dozens of pages long with her own diagrams.

This blog later became one of the most cited personal technical blogs in AI, used directly as textbook material by many universities.

In February 2025, she and former OpenAI CTO Mira Murati founded Thinking Machines Lab, with co-founders including OpenAI co-founder John Schulman, former research VPs Barret Zoph and Luke Metz. a16z led a $2B seed round at a $12B valuation.

And while her company was advancing rapidly, she took time to finish this long-form article on Scaling Laws, delayed for three years.

The ChatGPT, Claude, and Gemini you use every day are all governed by these formulas deciding how to train the next generation.

How good the next-generation AI is won't depend on who has more GPUs, but on who handles these details with greater precision.

References:

https://x.com/lilianweng/status/2070237256070389897?s=20

https://lilianweng.github.io/posts/2026-06-24-scaling-laws/

This article is from WeChat public account "新智元" (New AI Era), author: ASI启示录, editor: Moses

Criptos en tendencia

CitreaCTR

wrapped stUSDTWSTUSDT

Velodrome FinanceVELODROME

BrevisBREV

ZRX（0X）ZRX

PancakeSwapCAKE

Preguntas relacionadas

QWhat were the key differences in conclusions between OpenAI's Kaplan and DeepMind's Chinchilla regarding scaling laws and resource allocation?

AOpenAI's Kaplan (2020) concluded that for a given compute budget, the optimal model size should grow faster than the data, with the relationship N_opt ∝ C^0.73. This led to large, data-starved models like GPT-3. DeepMind's Chinchilla (2022) concluded the opposite: N_opt ∝ C^0.50, meaning model and data should scale proportionally. This led to the 'Chinchilla-optimal' 20:1 token-to-parameter ratio, and models trained this way (e.g., Chinchilla itself) outperformed larger, undertrained ones.

QWhat were the two main reasons identified for the contradictory findings between Kaplan and Chinchilla?

AThe contradictions stemmed from two main issues: 1) Parameter Counting: Kaplan excluded embedding parameters from his model size count (N), while Chinchilla included them. This difference in definition significantly affected the fitted scaling law exponent for smaller models. 2) Experimental Scale: Kaplan's experiments used models up to only ~1.5B parameters, while Chinchilla scaled up to ~16B+. The scaling exponent derived from a small-scale range can be misleading when extrapolated to much larger scales, where the exponent converges to a different value.

QWhat major methodological flaw was discovered in the Chinchilla paper years later, and what was its consequence?

AIn 2024, researchers found that Chinchilla's Method 3 (fitting the loss formula L(N, D)) had a critical flaw. The optimization used the mean of the Huber loss across samples instead of the sum. This resulted in a very small loss value, which caused the L-BFGS-B optimizer to stop prematurely, believing it had converged. Consequently, the published parameters were not the true global optimum. Additionally, key parameters were reported with insufficient precision (only 2 decimal places), leading to unrealistically narrow confidence intervals.

QWhat fundamental limitation of classical scaling laws is exposed by the impending exhaustion of high-quality text data?

AClassical scaling laws assume unique, non-repeating training data and an infinite supply of it. This premise is breaking down as high-quality text data is expected to be exhausted by 2026-2028. The industry now faces data repetition, which these original laws do not account for. Performance no longer scales predictably when models are trained on the same data for multiple epochs, leading to diminishing returns and overfitting.

QAccording to the article, what three main research directions is the AI industry pursuing to overcome the data limitation problem for scaling?

ATo bypass the data wall, the industry is focusing on three main research directions: 1) Reinforcement Learning: Using models to generate their own training signals through self-play on verifiable tasks (e.g., DeepSeek R1, OpenAI o-series). 2) Test-Time Computation: Enhancing model performance not during training, but by allowing it to 'think' for more steps (compute) when generating an answer (e.g., chain-of-thought). 3) Synthetic Data: Using existing powerful models to generate new data for training the next generation of models.

Lecturas Relacionadas

Assessing Sonic’s [S] 12% price drop and why more selling may be next

Sonic (S) has dropped 12% in 24 hours, with further declines anticipated. Despite a notable 23% surge in daily active users and over 228,000 on-chain transactions, key metrics suggest selling pressure is driving the activity. Total Value Locked (TVL) fell by $4.47 million to around $16.07 million, indicating capital outflows. While DEX volume spiked to $3.27 million and overall trading volume rose 51% to $26.8 million, this high volume alongside a falling price points to dominant selling momentum. Spot market data shows slight positive net inflows of about $20,000, hinting at minor accumulation, but this buying pressure is currently insufficient to counter the selling or fuel a near-term price recovery.

ambcryptoHace 50 min(s)

Assessing Sonic’s [S] 12% price drop and why more selling may be next

ambcryptoHace 50 min(s)

Stablecoins Becoming the Next Policy Challenge for the Fed's Walsh Version

Fed Governor Christopher Waller's speech at the June 22 conference on the U.S. dollar's international role signifies a notable policy shift: stablecoins like USDT and USDC are now being formally considered as potential channels for transmitting U.S. dollar liquidity globally. With their combined market cap surpassing $250 billion and high transaction volumes, these digital assets are moving from the periphery of crypto policy to the core of monetary system research. The key concern for policymakers is how stablecoin flows interact with traditional dollar infrastructure. Their growth could affect bank deposits, demand for short-term Treasury securities (like T-bills), and global access to dollars, depending on whether demand originates overseas or substitutes for domestic bank balances. Issuers' reserve management—holding assets in banks, money market funds, or Treasuries—links stablecoin activity directly to these core markets. The Fed's research agenda now examines whether stablecoins, by combining payment and balance-holding functions on digital rails, could complicate monetary policy implementation or transmit liquidity stress to banks. While current Treasury holdings by issuers are under 1% of the total market, their concentrated demand could marginally impact yields, especially during periods of stress. Consequently, stablecoins are evolving from mere crypto trading tools into a private-layer dollar transmission system with public policy implications, prompting closer regulatory scrutiny of their reserve robustness, redemption mechanisms, and systemic integration.

marsbitHace 1 hora(s)

Stablecoins Becoming the Next Policy Challenge for the Fed's Walsh Version

marsbitHace 1 hora(s)

After Burning Through 900 Billion, Zuckerberg Decides to Open a Casino Where You Can't Gamble

Mark Zuckerberg's Meta is venting a new app called Arena, a prediction market platform where users use points instead of real money to bet on outcomes of elections, sports, and world events. This move follows Meta's massive $90 billion cumulative loss in its Reality Labs metaverse division. Critics argue the "play money" model undermines the core mechanism of prediction markets, where real financial stakes ensure accurate forecasts. Arena is seen as Meta's attempt to copy established, fast-growing platforms like Polymarket and Kalshi while avoiding regulatory hurdles. The article suggests this may be less about creating a genuine prediction tool and more about social engagement, potentially acting as a gateway that introduces billions to the concept, from which licensed platforms could ultimately benefit.

marsbitHace 1 hora(s)

After Burning Through 900 Billion, Zuckerberg Decides to Open a Casino Where You Can't Gamble

marsbitHace 1 hora(s)

A 380% Soar, Shenzhen’s 100-Billion-Yuan IPO Rings the Bell

HKC Holdings, a major Chinese display panel manufacturer, has successfully listed on the Shenzhen Stock Exchange's main board. The company's shares surged over 380% on its debut, pushing its market capitalization to around 350 billion yuan (formerly reaching 500 billion yuan). Founded by Wang Zhiyong in Shenzhen's Huaqiangbei electronics market nearly three decades ago, HKC evolved from assembling monitors to becoming a global top-tier supplier of semiconductor display panels for TVs, monitors, and smartphones. The IPO marks a significant milestone for HKC and its backers. The company's growth into the capital-intensive panel manufacturing sector was supported through partnerships with state-owned capital from regions like Chongqing, Mianyang, and Chuzhou. Its shareholder list also includes BOE Technology's investment arm. In recent years, HKC reported strong financials, with core panel business contributing over 70% of revenue and clients including Samsung, TCL, and Xiaomi. This listing is seen as part of a broader trend in Shenzhen's evolving tech landscape. Beyond established giants, the city is nurturing clusters of leading companies in specialized sectors like robotics—exemplified by the "Shenzhen Robot Valley"—and storage chips, where a group of firms dubbed the "Storage Five Tigers" has achieved a combined trillion-yuan market valuation. Shenzhen's strategic focus on emerging industries such as AI terminals, low-altitude economy, and humanoid robotics aims to build new industrial depth and foster the next generation of tech champions.

marsbitHace 1 hora(s)

A 380% Soar, Shenzhen’s 100-Billion-Yuan IPO Rings the Bell

marsbitHace 1 hora(s)

Lasting 233 Days with Over 50% Drawdown, Is This the Mildest Bear Market to Date?

Bitcoin's current bear market has lasted 233 days as of June 24, making it the fourth-longest cycle since 2014, defined by its price staying below the 200-day moving average (200 DMA) for over 30 days. Historically, the longest bear markets were in 2018-2019 (385 days) and 2022-2023 (381 days), driven by major structural crashes post-all-time highs. In contrast, the current 2025-2026 downturn, following Bitcoin's January 2025 peak of $124,773, appears driven by broader macro shifts like rising interest rates and fading post-halving momentum. This cycle is notably the mildest on record so far, with a peak-to-trough drawdown of 51.2%. Previous major bear markets saw declines between 76.7% and 83.6%. Even the sharp but brief 2020 COVID crash caused a 74.4% drop. As of the analysis date, Bitcoin's price is $62,651, about 22% below its 200 DMA of $76,450. A sustained rally of over one-fifth is needed to reclaim this key level. If the recent low of $60,861 on June 7, 2026, is confirmed as the cycle bottom, historical recovery patterns suggest reclaiming the 200 DMA could take from 65 to 166 days, potentially placing it around August 2026 at the earliest.

marsbitHace 1 hora(s)

Lasting 233 Days with Over 50% Drawdown, Is This the Mildest Bear Market to Date?

marsbitHace 1 hora(s)

Trading

Spot

Futuros

Artículos destacados

Qué es $S$

Entendiendo SPERO: Una Visión General Completa Introducción a SPERO A medida que el panorama de la innovación continúa evolucionando, la aparición de tecnologías web3 y proyectos de criptomonedas juega un papel fundamental en la configuración del futuro digital. Un proyecto que ha atraído la atención en este campo dinámico es SPERO, denotado como SPERO,$$s$. Este artículo tiene como objetivo reunir y presentar información detallada sobre SPERO, para ayudar a entusiastas e inversores a comprender sus fundamentos, objetivos e innovaciones dentro de los dominios web3 y cripto. ¿Qué es SPERO,$$s$? SPERO,$$s$ es un proyecto único dentro del espacio cripto que busca aprovechar los principios de descentralización y tecnología blockchain para crear un ecosistema que promueva la participación, la utilidad y la inclusión financiera. El proyecto está diseñado para facilitar interacciones de igual a igual de nuevas maneras, proporcionando a los usuarios soluciones y servicios financieros innovadores. En su esencia, SPERO,$$s$ tiene como objetivo empoderar a los individuos al proporcionar herramientas y plataformas que mejoren la experiencia del usuario en el espacio de las criptomonedas. Esto incluye habilitar métodos de transacción más flexibles, fomentar iniciativas impulsadas por la comunidad y crear caminos para oportunidades financieras a través de aplicaciones descentralizadas (dApps). La visión subyacente de SPERO,$$s$ gira en torno a la inclusividad, buscando cerrar brechas dentro de las finanzas tradicionales mientras aprovecha los beneficios de la tecnología blockchain. ¿Quién es el Creador de SPERO,$$s$? La identidad del creador de SPERO,$$s$ sigue siendo algo oscura, ya que hay recursos públicos limitados que proporcionan información de fondo detallada sobre su(s) fundador(es). Esta falta de transparencia puede derivarse del compromiso del proyecto con la descentralización, una ética que muchos proyectos web3 comparten, priorizando las contribuciones colectivas sobre el reconocimiento individual. Al centrar las discusiones en torno a la comunidad y sus objetivos colectivos, SPERO,$$s$ encarna la esencia del empoderamiento sin señalar a individuos específicos. Como tal, comprender la ética y la misión de SPERO sigue siendo más importante que identificar a un creador singular. ¿Quiénes son los Inversores de SPERO,$$s$? SPERO,$$s$ cuenta con el apoyo de una diversa gama de inversores que van desde capitalistas de riesgo hasta inversores ángeles dedicados a fomentar la innovación en el sector cripto. El enfoque de estos inversores generalmente se alinea con la misión de SPERO, priorizando proyectos que prometen avances tecnológicos sociales, inclusión financiera y gobernanza descentralizada. Estas fundaciones de inversores suelen estar interesadas en proyectos que no solo ofrecen productos innovadores, sino que también contribuyen positivamente a la comunidad blockchain y sus ecosistemas. El respaldo de estos inversores refuerza a SPERO,$$s$ como un contendiente notable en el dominio de proyectos cripto que evoluciona rápidamente. ¿Cómo Funciona SPERO,$$s$? SPERO,$$s$ emplea un marco multifacético que lo distingue de los proyectos de criptomonedas convencionales. Aquí hay algunas de las características clave que subrayan su singularidad e innovación: Gobernanza Descentralizada: SPERO,$$s$ integra modelos de gobernanza descentralizada, empoderando a los usuarios para participar activamente en los procesos de toma de decisiones sobre el futuro del proyecto. Este enfoque fomenta un sentido de propiedad y responsabilidad entre los miembros de la comunidad. Utilidad del Token: SPERO,$$s$ utiliza su propio token de criptomoneda, diseñado para servir diversas funciones dentro del ecosistema. Estos tokens permiten transacciones, recompensas y la facilitación de servicios ofrecidos en la plataforma, mejorando la participación y la utilidad general. Arquitectura en Capas: La arquitectura técnica de SPERO,$$s$ apoya la modularidad y escalabilidad, permitiendo la integración fluida de características y aplicaciones adicionales a medida que el proyecto evoluciona. Esta adaptabilidad es fundamental para mantener la relevancia en el cambiante paisaje cripto. Participación de la Comunidad: El proyecto enfatiza iniciativas impulsadas por la comunidad, empleando mecanismos que incentivan la colaboración y la retroalimentación. Al nutrir una comunidad sólida, SPERO,$$s$ puede abordar mejor las necesidades de los usuarios y adaptarse a las tendencias del mercado. Enfoque en la Inclusión: Al ofrecer tarifas de transacción bajas e interfaces amigables para el usuario, SPERO,$$s$ busca atraer a una base de usuarios diversa, incluyendo a individuos que anteriormente pueden no haber participado en el espacio cripto. Este compromiso con la inclusión se alinea con su misión general de empoderamiento a través de la accesibilidad. Cronología de SPERO,$$s$ Entender la historia de un proyecto proporciona información crucial sobre su trayectoria de desarrollo y hitos. A continuación se presenta una cronología sugerida que mapea eventos significativos en la evolución de SPERO,$$s$: Fase de Conceptualización e Ideación: Las ideas iniciales que forman la base de SPERO,$$s$ fueron concebidas, alineándose estrechamente con los principios de descentralización y enfoque comunitario dentro de la industria blockchain. Lanzamiento del Whitepaper del Proyecto: Tras la fase conceptual, se lanzó un whitepaper completo que detalla la visión, los objetivos y la infraestructura tecnológica de SPERO,$$s$ para generar interés y retroalimentación de la comunidad. Construcción de Comunidad y Primeras Interacciones: Se realizaron esfuerzos de divulgación activa para construir una comunidad de primeros adoptantes y posibles inversores, facilitando discusiones en torno a los objetivos del proyecto y obteniendo apoyo. Evento de Generación de Tokens: SPERO,$$s$ llevó a cabo un evento de generación de tokens (TGE) para distribuir sus tokens nativos a los primeros seguidores y establecer liquidez inicial dentro del ecosistema. Lanzamiento de la dApp Inicial: La primera aplicación descentralizada (dApp) asociada con SPERO,$$s$ se puso en marcha, permitiendo a los usuarios interactuar con las funcionalidades centrales de la plataforma. Desarrollo Continuo y Alianzas: Actualizaciones y mejoras continuas a las ofertas del proyecto, incluyendo alianzas estratégicas con otros actores en el espacio blockchain, han moldeado a SPERO,$$s$ en un jugador competitivo y en evolución en el mercado cripto. Conclusión SPERO,$$s$ se erige como un testimonio del potencial de web3 y las criptomonedas para revolucionar los sistemas financieros y empoderar a los individuos. Con un compromiso con la gobernanza descentralizada, la participación comunitaria y funcionalidades diseñadas de manera innovadora, allana el camino hacia un paisaje financiero más inclusivo. Como con cualquier inversión en el espacio cripto que evoluciona rápidamente, se anima a los posibles inversores y usuarios a investigar a fondo y participar de manera reflexiva con los desarrollos en curso dentro de SPERO,$$s$. El proyecto muestra el espíritu innovador de la industria cripto, invitando a una mayor exploración de sus innumerables posibilidades. Mientras el viaje de SPERO,$$s$ aún se desarrolla, sus principios fundamentales pueden, de hecho, influir en el futuro de cómo interactuamos con la tecnología, las finanzas y entre nosotros en ecosistemas digitales interconectados.

88 Vistas totalesPublicado en 2024.12.17Actualizado en 2024.12.17

Qué es AGENT S

Agent S: El Futuro de la Interacción Autónoma en Web3 Introducción En el paisaje en constante evolución de Web3 y las criptomonedas, las innovaciones están redefiniendo constantemente cómo los individuos interactúan con las plataformas digitales. Uno de estos proyectos pioneros, Agent S, promete revolucionar la interacción humano-computadora a través de su marco agente abierto. Al allanar el camino para interacciones autónomas, Agent S busca simplificar tareas complejas, ofreciendo aplicaciones transformadoras en inteligencia artificial (IA). Esta exploración detallada profundizará en las complejidades del proyecto, sus características únicas y las implicaciones para el dominio de las criptomonedas. ¿Qué es Agent S? Agent S se presenta como un marco agente abierto innovador, diseñado específicamente para abordar tres desafíos fundamentales en la automatización de tareas informáticas: Adquisición de Conocimiento Específico del Dominio: El marco aprende inteligentemente de diversas fuentes de conocimiento externas y experiencias internas. Este enfoque dual le permite construir un rico repositorio de conocimiento específico del dominio, mejorando su rendimiento en la ejecución de tareas. Planificación a Largo Plazo de Tareas: Agent S emplea planificación jerárquica aumentada por la experiencia, un enfoque estratégico que facilita la descomposición y ejecución eficiente de tareas complejas. Esta característica mejora significativamente su capacidad para gestionar múltiples subtareas de manera eficiente y efectiva. Manejo de Interfaces Dinámicas y No Uniformes: El proyecto introduce la Interfaz Agente-Computadora (ACI), una solución innovadora que mejora la interacción entre agentes y usuarios. Utilizando Modelos de Lenguaje Multimodal de Gran Escala (MLLMs), Agent S puede navegar y manipular diversas interfaces gráficas de usuario sin problemas. A través de estas características pioneras, Agent S proporciona un marco robusto que aborda las complejidades involucradas en la automatización de la interacción humana con las máquinas, preparando el terreno para una multitud de aplicaciones en IA y más allá. ¿Quién es el Creador de Agent S? Si bien el concepto de Agent S es fundamentalmente innovador, la información específica sobre su creador sigue siendo elusiva. El creador es actualmente desconocido, lo que resalta ya sea la etapa incipiente del proyecto o la elección estratégica de mantener a los miembros fundadores en el anonimato. Independientemente de la anonimidad, el enfoque sigue siendo en las capacidades y el potencial del marco. ¿Quiénes son los Inversores de Agent S? Dado que Agent S es relativamente nuevo en el ecosistema criptográfico, la información detallada sobre sus inversores y patrocinadores financieros no está documentada explícitamente. La falta de información disponible públicamente sobre las bases de inversión u organizaciones que apoyan el proyecto plantea preguntas sobre su estructura de financiamiento y hoja de ruta de desarrollo. Comprender el respaldo es crucial para evaluar la sostenibilidad del proyecto y su posible impacto en el mercado. ¿Cómo Funciona Agent S? En el núcleo de Agent S se encuentra una tecnología de vanguardia que le permite funcionar de manera efectiva en diversos entornos. Su modelo operativo se basa en varias características clave: Interacción Humano-Computadora Similar a la Humana: El marco ofrece planificación avanzada de IA, esforzándose por hacer que las interacciones con las computadoras sean más intuitivas. Al imitar el comportamiento humano en la ejecución de tareas, promete elevar las experiencias de los usuarios. Memoria Narrativa: Empleada para aprovechar experiencias de alto nivel, Agent S utiliza memoria narrativa para hacer un seguimiento de las historias de tareas, mejorando así sus procesos de toma de decisiones. Memoria Episódica: Esta característica proporciona a los usuarios una guía paso a paso, permitiendo que el marco ofrezca apoyo contextual a medida que se desarrollan las tareas. Soporte para OpenACI: Con la capacidad de ejecutarse localmente, Agent S permite a los usuarios mantener el control sobre sus interacciones y flujos de trabajo, alineándose con la ética descentralizada de Web3. Fácil Integración con APIs Externas: Su versatilidad y compatibilidad con varias plataformas de IA aseguran que Agent S pueda encajar sin problemas en ecosistemas tecnológicos existentes, convirtiéndolo en una opción atractiva para desarrolladores y organizaciones. Estas funcionalidades contribuyen colectivamente a la posición única de Agent S dentro del espacio cripto, ya que automatiza tareas complejas y de múltiples pasos con una intervención humana mínima. A medida que el proyecto evoluciona, sus posibles aplicaciones en Web3 podrían redefinir cómo se desarrollan las interacciones digitales. Cronología de Agent S El desarrollo y los hitos de Agent S pueden encapsularse en una cronología que resalta sus eventos significativos: 27 de septiembre de 2024: El concepto de Agent S fue lanzado en un documento de investigación integral titulado “Un Marco Agente Abierto que Usa Computadoras Como un Humano”, mostrando las bases del proyecto. 10 de octubre de 2024: El documento de investigación fue puesto a disposición del público en arXiv, ofreciendo una exploración profunda del marco y su evaluación de rendimiento basada en el benchmark OSWorld. 12 de octubre de 2024: Se lanzó una presentación en video, proporcionando una visión visual de las capacidades y características de Agent S, involucrando aún más a posibles usuarios e inversores. Estos marcadores en la cronología no solo ilustran el progreso de Agent S, sino que también indican su compromiso con la transparencia y la participación comunitaria. Puntos Clave Sobre Agent S A medida que el marco Agent S continúa evolucionando, varios atributos clave destacan, subrayando su naturaleza innovadora y potencial: Marco Innovador: Diseñado para proporcionar un uso intuitivo de las computadoras similar a la interacción humana, Agent S aporta un enfoque novedoso a la automatización de tareas. Interacción Autónoma: La capacidad de interactuar de manera autónoma con las computadoras a través de GUI significa un salto hacia soluciones informáticas más inteligentes y eficientes. Automatización de Tareas Complejas: Con su metodología robusta, puede automatizar tareas complejas y de múltiples pasos, haciendo que los procesos sean más rápidos y menos propensos a errores. Mejora Continua: Los mecanismos de aprendizaje permiten a Agent S mejorar a partir de experiencias pasadas, mejorando continuamente su rendimiento y eficacia. Versatilidad: Su adaptabilidad en diferentes entornos operativos como OSWorld y WindowsAgentArena asegura que pueda servir a una amplia gama de aplicaciones. A medida que Agent S se posiciona en el paisaje de Web3 y criptomonedas, su potencial para mejorar las capacidades de interacción y automatizar procesos significa un avance significativo en las tecnologías de IA. A través de su marco innovador, Agent S ejemplifica el futuro de las interacciones digitales, prometiendo una experiencia más fluida y eficiente para los usuarios en diversas industrias. Conclusión Agent S representa un audaz avance en la unión de la IA y Web3, con la capacidad de redefinir cómo interactuamos con la tecnología. Aunque aún se encuentra en sus primeras etapas, las posibilidades para su aplicación son vastas y atractivas. A través de su marco integral que aborda desafíos críticos, Agent S busca llevar las interacciones autónomas al primer plano de la experiencia digital. A medida que nos adentramos más en los reinos de las criptomonedas y la descentralización, proyectos como Agent S sin duda desempeñarán un papel crucial en la configuración del futuro de la tecnología y la colaboración humano-computadora.

500 Vistas totalesPublicado en 2025.01.14Actualizado en 2025.01.14

Cómo comprar S

¡Bienvenido a HTX.com! Hemos hecho que comprar Sonic (S) sea simple y conveniente. Sigue nuestra guía paso a paso para iniciar tu viaje de criptos.Paso 1: crea tu cuenta HTXUtiliza tu correo electrónico o número de teléfono para registrarte y obtener una cuenta gratuita en HTX. Experimenta un proceso de registro sin complicaciones y desbloquea todas las funciones.Obtener mi cuentaPaso 2: ve a Comprar cripto y elige tu método de pagoTarjeta de crédito/débito: usa tu Visa o Mastercard para comprar Sonic (S) al instante.Saldo: utiliza fondos del saldo de tu cuenta HTX para tradear sin problemas.Terceros: hemos agregado métodos de pago populares como Google Pay y Apple Pay para mejorar la comodidad.P2P: tradear directamente con otros usuarios en HTX.Over-the-Counter (OTC): ofrecemos servicios personalizados y tipos de cambio competitivos para los traders.Paso 3: guarda tu Sonic (S)Después de comprar tu Sonic (S), guárdalo en tu cuenta HTX. Alternativamente, puedes enviarlo a otro lugar mediante transferencia blockchain o utilizarlo para tradear otras criptomonedas.Paso 4: tradear Sonic (S)Tradear fácilmente con Sonic (S) en HTX's mercado spot. Simplemente accede a tu cuenta, selecciona tu par de trading, ejecuta tus trades y monitorea en tiempo real. Ofrecemos una experiencia fácil de usar tanto para principiantes como para traders experimentados.

1.0k Vistas totalesPublicado en 2025.01.15Actualizado en 2026.06.02

Discusiones

Bienvenido a la comunidad de HTX. Aquí puedes mantenerte informado sobre los últimos desarrollos de la plataforma y acceder a análisis profesionales del mercado. A continuación se presentan las opiniones de los usuarios sobre el precio de S (S).