From 32 items, 10 important content pieces were selected

1. Moonshot Releases 2.8T Parameter Kimi K3 Open-Weight Model ⭐️ 9.0/10
2. Coinbase to Suspend ACX Trading July 28, 2026 ⭐️ 8.0/10
3. First Implicit Tactile World Action Model Trained on 500k Hours of Video ⭐️ 8.0/10
4. World's First Agentic Diffusion Model with 128K Context ⭐️ 8.0/10
5. A $500 RL Fine-Tune of 9B Open Model Beats Frontier Models ⭐️ 8.0/10
6. Anthropic Publishes Stance on Open-Weights AI Models ⭐️ 8.0/10
7. Documentation for self-contained Python builds ⭐️ 8.0/10
8. Jensen Huang bets $5 billion on Ilya Sutskever's AI scaling approach ⭐️ 7.0/10
9. Kalshi, Polymarket win pause against Minnesota prediction market ban ⭐️ 7.0/10
10. Crypto's $90 Trillion Repo Market Comes to Wall Street, Slowly ⭐️ 7.0/10

Moonshot Releases 2.8T Parameter Kimi K3 Open-Weight Model ⭐️ 9.0/10

Moonshot has released the weights for their 2.8 trillion parameter Kimi K3 model on Hugging Face, with the 1.56TB checkpoint now available under a new license that restricts commercial use by large Model-as-a-Service providers. OpenRouter already offers K3 from seven providers at the same pricing as Moonshot. This release represents a major milestone in open-weight AI, as a 2.8 trillion parameter model is among the largest ever made publicly available. The shift from a modified MIT license to a more restrictive one for large commercial entities may set a precedent for how future open-weight models balance openness with commercial protection. The new license for K3 requires Model-as-a-Service businesses exceeding $20 million in annual revenue to enter a separate agreement with Moonshot, whereas previous K2 license only required attribution for entities with over 100 million MAUs or $20M monthly revenue. Moonshot consistently uses the term 'open weight' rather than 'open source' to describe this release.

rss · Simon Willison · Jul 28, 07:39

Background: Model weights are the learned parameters that define how an AI model processes inputs to produce outputs, typically resulting from training on large datasets. In large language models, parameters include billions or trillions of adjustable values that encode the model's knowledge and behavior. Open-weight models allow users to download and run the weights locally, but may come with usage restrictions that distinguish them from fully open-source software.

Coinbase to Suspend ACX Trading July 28, 2026 ⭐️ 8.0/10

Coinbase announced it will suspend trading of the ACX token on July 28, 2026, as the project team is winding down the token and offering conversion options to holders. The exchange has already moved ACX order books to limit-only mode ahead of the suspension. This suspension removes a major trading venue for ACX, affecting liquidity and access for holders who wish to trade or exit their positions. It highlights the risks of token wind-downs and the importance of following project communications for conversion opportunities. Trading on Coinbase will be suspended from July 28 to August 3, 2026, and holders must move their ACX to a self-custodial wallet to participate in the project's conversion options. The ACX token is used for governance, staking, and fee payments within the Across Protocol ecosystem.

rss · Coinbase Status · Jul 28, 15:00

Background: The ACX token is the native token of Across Protocol, a cross-chain bridge that allows token transfers between different blockchains. It is used for governance voting, staking to become a relayer, and paying transaction fees. A token wind-down occurs when a project decides to discontinue its token, often offering holders a conversion to another asset or stablecoins. Limit-only mode means traders can only place limit orders, not market orders, which is a common step before a full trading suspension.

First Implicit Tactile World Action Model Trained on 500k Hours of Video ⭐️ 8.0/10

Researchers have trained the first implicit tactile world action model using 500,000 hours of video, enabling AI systems to understand and simulate tactile interactions. This marks a significant step in embodied AI and multimodal learning, allowing robots to manipulate objects and interact with the environment more naturally, with profound implications for automation, healthcare, and home robotics. The model employs implicit representation learning, likely combining world models with tactile sensing data, and was trained on an enormous dataset of 500,000 hours of video. Specific architectural details have not been publicly disclosed, but this is the first attempt to integrate tactile sensing into a world model.

rss · QbitAI 量子位 · Jul 28, 13:25

Background: A world model is an internal simulation of the environment that allows AI to predict consequences of actions. Tactile sensing technology aims to replicate human touch. Previous world models relied primarily on visual and textual inputs, lacking the tactile modality.

Tags: #AI, #robotics, #world model, #tactile sensing, #machine learning

World's First Agentic Diffusion Model with 128K Context ⭐️ 8.0/10

Researchers have released the first agentic diffusion model capable of real-time action and error correction, achieving a 128K context length that matches autoregressive models. This breakthrough extends diffusion models to long-horizon agent tasks, potentially enabling more robust and interactive AI agents that can plan and adapt without full retraining. The model supports 128K token context, matching the context window of leading autoregressive models, and incorporates an agentic loop that iteratively refines outputs.

rss · QbitAI 量子位 · Jul 28, 12:18

Background: Diffusion models traditionally excel at generating high-quality images and audio by gradually denoising random noise. Autoregressive models have dominated long-context tasks like text generation and planning. This work bridges the gap by enabling diffusion models to handle agentic tasks requiring step-by-step decision-making and error correction.

Tags: #扩散模型, #Agent, #长上下文, #AI研究, #突破

A $500 RL Fine-Tune of 9B Open Model Beats Frontier Models ⭐️ 8.0/10

A team fine-tuned a 9 billion parameter open-weight model using reinforcement learning for under $500, achieving better performance on catalog review tasks than much larger frontier models like GPT-4 or Claude. This demonstrates that small, fine-tuned models can be highly cost-effective for specific tasks, challenging the economic rationale for massive frontier models and the associated infrastructure spending. The fine-tuning process utilized reinforcement learning to optimize the model for catalog review accuracy. The specific model, dataset, and exact training cost breakdown were detailed in the blog post, though the author notes potential overfitting concerns.

hackernews · ilreb · Jul 28, 10:18 · Discussion

Background: Reinforcement learning fine-tuning (RLFT) is a technique where a language model is trained to maximize a reward signal based on its outputs, aligning it with specific goals. Catalog review involves automatically evaluating product descriptions, metadata, or listings for quality and consistency. Frontier models like GPT-4 and Claude are very large, general-purpose models that are expensive to train and run.

Discussion: Commenters largely praised the cost-efficiency demonstration, with some noting that most use cases don't need massive models. However, concerns were raised about overfitting due to lack of held-out test data and that such specialized benchmarks don't accurately compare to frontier models' generalized capabilities.

Tags: #AI, #fine-tuning, #open-source, #cost-efficiency, #RL

Anthropic Publishes Stance on Open-Weights AI Models ⭐️ 8.0/10

Anthropic published an official statement outlining its position on open-weights AI models, advocating for mandatory safety testing for all sufficiently capable models rather than a blanket ban. The statement clarifies that Anthropic has never called for banning open-weights models but emphasizes the need for regulation as capabilities advance. This statement fuels the ongoing debate between open and closed AI development, as open-weights models lower barriers to access but raise safety concerns. Anthropic's position could influence regulatory policies, especially as governments consider how to balance innovation and risk. Anthropic's CEO Dario Amodei wrote the statement, which also supports banning chip sales to China and cracking down on smuggling of chips to prevent advanced AI capabilities from proliferating. The statement explicitly opposes a ban on open-weights models but supports mandatory safety testing that critics argue could effectively function as a ban.

hackernews · surprisetalk · Jul 28, 06:03 · Discussion

Background: Open-weights AI models are neural network models whose trained parameters (weights and biases) are publicly released for anyone to download, modify, and run. This contrasts with closed models where only an API is accessible. The open-weights approach enables wider access and customization but also raises concerns about misuse, as developers cannot control how the model is used after release.

Discussion: Commenters largely criticized Anthropic for perceived hypocrisy, pointing out contradictions between supporting chip bans while opposing open-weights bans, and questioning the feasibility of safety testing regimes. Some accused Anthropic of virtue signaling and protecting its commercial interests under the guise of safety concerns.

Documentation for self-contained Python builds ⭐️ 8.0/10

Documentation for python-build-standalone, a set of self-contained portable Python distributions, has been published, detailing how these builds are used by tools like uv, pipx, and Hatch to provide Python installations without system dependencies. This infrastructure simplifies Python deployment across platforms, enables bundling Python into desktop applications, and reduces friction for users who need a specific Python version without system-level installation. The distributions are maintained by Astral (the company behind uv), support many platforms (Linux, macOS, Windows), and are available for multiple Python versions. They are designed to be fully relocatable and have no external dependencies.

hackernews · jcbhmr · Jul 28, 02:43 · Discussion

Background: Standard Python distributions are often tied to a specific operating system or require compiling from source. python-build-standalone provides pre-built, relocatable Python binaries that can be extracted and used anywhere, making them ideal for packaging tools like package managers or embedding in applications. They are produced using a custom build process that strips out system-specific baggage.

Discussion: Comments highlight widespread adoption: charliermarsh (uv lead) confirms uv uses these distributions and notes significant engineering effort in keeping up with CPython. simonw praises them as excellent for bundling Python into apps, and zie points out an alternative approach using Cosmopolitan cross-platform binaries. rsyring mentions PyOxy as a sister project that creates single-file executables from these distributions.

Jensen Huang bets $5 billion on Ilya Sutskever's AI scaling approach ⭐️ 7.0/10

NVIDIA CEO Jensen Huang has invested $5 billion in Ilya Sutskever's AI scaling approach, signaling a major bet on continuing to scale compute for AI models. This investment is significant because it challenges recent skepticism about the limits of scaling and reaffirms the industry's commitment to scaling laws as a path to more capable AI systems. The $5 billion investment is directed at Ilya Sutskever's current venture, Safe Superintelligence Inc., which focuses on safe AI development. Jensen Huang's bet indicates a belief that scaling compute remains crucial for AI progress.

rss · QbitAI 量子位 · Jul 28, 13:46

Background: Neural scaling laws describe how neural network performance improves as model size, dataset size, and compute are scaled up. Ilya Sutskever, co-founder of OpenAI, previously suggested that the era of scaling compute alone was ending and that AI research needed to explore new approaches. This investment signals a possible shift back to scaling, or a hybrid strategy.

Tags: #AI, #Scaling, #Investment, #Ilya Sutskever, #Jensen Huang

Kalshi, Polymarket win pause against Minnesota prediction market ban ⭐️ 7.0/10

A federal judge granted a preliminary injunction pausing enforcement of Minnesota's ban on prediction markets, allowing Kalshi and Polymarket to continue operating in the state while the legal challenge proceeds. This ruling sets a precedent for prediction market regulation in the U.S., potentially influencing other states' approaches and underscoring the legal distinction between prediction markets and gambling. Minnesota had banned prediction markets as unlawful gambling, but the court found the platforms likely to succeed on First Amendment grounds, as prediction markets involve expressive conduct protected under commercial speech.

rss · CoinDesk · Jul 28, 07:29

Background: Kalshi is a regulated prediction market exchange based in New York, while Polymarket is a decentralized platform built on the Polygon blockchain. Prediction markets allow users to trade contracts based on the outcome of future events, such as elections or sports, and are considered by some to be a form of expressive activity rather than gambling.

Crypto's $90 Trillion Repo Market Comes to Wall Street, Slowly ⭐️ 7.0/10

Traditional finance's tri-party repo market, valued at $90 trillion, is being adapted for cryptocurrency trading, but major banks are proceeding cautiously. This development signals growing institutional interest in crypto, as the repo market is a critical source of short-term funding and liquidity. However, the slow adoption by big banks indicates that crypto market infrastructure is not yet fully integrated with traditional finance. Tri-party repo involves a third-party agent handling collateral management, reducing operational burden. The adaptation for crypto could involve tokenized collateral and blockchain-based settlement.

rss · CoinDesk · Jul 28, 02:00

Background: The repo market is a short-term secured lending market where one party sells securities to another and agrees to repurchase them later at a higher price. Tri-party repo outsources post-trade processing to a third-party agent, making it more efficient. The crypto adaptation could enable institutions to use digital assets as collateral for cash loans.

From 130 items, 10 important content pieces were selected

1. Thailand SEC Files Criminal Charges Against Bitkub Over $47M Hack Cover-Up ⭐️ 9.0/10
2. Lido Moves $16.5B in Staked ETH to Slash Validators by a Third ⭐️ 8.0/10
3. Circle to acquire nearly 1,000 IBM blockchain patents ⭐️ 8.0/10
4. Bun's Rust Rewrite Progresses, Delays v1.4 for Node.js Tests ⭐️ 8.0/10
5. Triple-A confirms treasury-wallet breach, $11.8M loss ⭐️ 8.0/10
6. BitMart withdrawals slow after wind-down announcement ⭐️ 8.0/10
7. Storj files for Chapter 11 bankruptcy, explores equity for tokenholders ⭐️ 8.0/10
8. Binance Announces Delisting of 5 Altcoins ⭐️ 7.0/10
9. go-ethereum v1.17.5: Maintenance Release with Pebble v2 and Amsterdam EIPs ⭐️ 7.0/10
10. US citizen charged after GrapheneOS phone wipe at border ⭐️ 7.0/10

Thailand SEC Files Criminal Charges Against Bitkub Over $47M Hack Cover-Up ⭐️ 9.0/10

Thailand's Securities and Exchange Commission has filed a criminal complaint against cryptocurrency exchange Bitkub and two former directors for failing to disclose a $47 million hack that occurred in 2021. This marks a significant regulatory enforcement action that underscores the importance of transparency and honesty in the crypto industry, potentially affecting user trust and setting a precedent for other exchanges. The SEC alleges that Bitkub omitted the $47 million theft from its filings, and the complaint targets both the company and two former directors personally, indicating a focus on individual accountability.

rss · Decrypt · Jul 27, 17:16

Background: Bitkub is a Thailand-based cryptocurrency exchange founded in 2018, and it was among the first to receive a digital asset license from Thailand's SEC in 2019. The exchange has been a major player in the Thai crypto market. The SEC is the primary regulator for digital assets in Thailand, responsible for enforcing securities laws.

Lido Moves $16.5B in Staked ETH to Slash Validators by a Third ⭐️ 8.0/10

Lido has begun moving approximately $16.5 billion worth of staked ether (stETH) to consolidate its validator operations, aiming to reduce the total number of validators it runs by about one-third. This move could improve capital efficiency and reduce operational costs for Lido, the largest Ethereum liquid staking protocol, potentially altering staking rewards distribution and affecting decentralization dynamics in Ethereum's staking ecosystem. The consolidation involves merging smaller validator deposits into larger ones, as each Ethereum validator requires a minimum of 32 ETH to activate; reducing validator count may streamline rewards and lower overhead for Lido node operators.

rss · CoinDesk · Jul 27, 22:00

Background: Lido is a liquid staking protocol that allows users to stake any amount of ETH and receive stETH, a tradable token representing their staked position. Ethereum's proof-of-stake consensus requires validators to lock 32 ETH to secure the network and earn rewards. Lido pools user deposits to run validators on their behalf, and this consolidation aims to optimize its validator fleet.

Tags: #Lido, #Ethereum, #Staking, #DeFi, #Protocol Update

Circle to acquire nearly 1,000 IBM blockchain patents ⭐️ 8.0/10

Circle, the issuer of the USDC stablecoin, announced it will acquire nearly 1,000 blockchain patents from IBM, including over 680 patent families. This strategic move strengthens Circle's intellectual property portfolio in blockchain technology. This acquisition positions Circle as a major holder of blockchain patents, potentially creating a competitive moat in the stablecoin and blockchain infrastructure space. It signals long-term consolidation and investment in intellectual property by key industry players, which could shape future standards and licensing. IBM's blockchain portfolio focuses on supply chain applications, with nearly 1,000 issued patents worldwide. The deal's financial terms were not disclosed, and it is subject to regulatory approvals.

rss · Cointelegraph · Jul 27, 21:08

Background: Circle is the company behind USDC, a stablecoin pegged 1:1 to the US dollar and backed by reserves of cash and short-term US Treasury bonds. IBM has been a pioneer in enterprise blockchain, developing networks for food safety, shipping, trade finance, and more. Patents are crucial for protecting and monetizing blockchain innovations.

Tags: #Circle, #USDC, #IBM, #patents, #blockchain

Bun's Rust Rewrite Progresses, Delays v1.4 for Node.js Tests ⭐️ 8.0/10

Bun's creator Jarred announced that the Rust rewrite has shipped in Claude Code over a month ago, but the v1.4 release is delayed until a promised number of Node.js compatibility tests pass, with a target of next Tuesday. This update provides transparency on a major technical overhaul of Bun, a popular JavaScript runtime, and highlights the challenges of achieving Node.js compatibility after a full rewrite. The Rust rewrite was already used in Claude Code, but Bun v1.4 is delayed because a specific number of newly passing Node.js tests has not been met; the pull requests are pending and merging is expected soon.

hackernews · tomlockwood · Jul 27, 19:12 · Discussion

Background: Bun is a high-performance JavaScript runtime that aims to be a drop-in replacement for Node.js, written in Zig. The project announced a rewrite in Rust to improve performance and safety. The rewrite has been shipped in Claude Code, an AI coding tool, but the general release is pending further compatibility improvements.

Discussion: Community members expressed mixed feelings: some saw the delay as a normal part of a major refactor, while others questioned the premature claims of success from LLM-assisted translation. There was also curiosity about the incomplete discussion thread.

Tags: #bun, #rust, #javascript-runtime, #rewrite, #nodejs-compatibility

Triple-A confirms treasury-wallet breach, $11.8M loss ⭐️ 8.0/10

Triple-A, a stablecoin payments company, confirmed a breach of its treasury wallet resulting in $11.8 million in losses. The company stated that client funds were unaffected and the financial impact will be covered by its treasury reserves. This incident highlights ongoing security vulnerabilities in crypto payment infrastructure, potentially impacting trust in stablecoin-based payment systems. However, the assurance that client funds are safe may mitigate immediate panic among users. The breach specifically targeted Triple-A's treasury wallet, not client wallets. The company plans to absorb the $11.8M loss through its own reserves, indicating a degree of financial resilience.

rss · Cointelegraph · Jul 27, 13:15

Background: Triple-A is a stablecoin payments company that facilitates crypto-to-fiat transactions for businesses. Treasury wallets are used by companies to manage operational funds, separate from client funds. This breach underscores the importance of robust security measures for corporate crypto wallets.

BitMart withdrawals slow after wind-down announcement ⭐️ 8.0/10

BitMart users are experiencing withdrawal delays and freeze notices as the exchange winds down, with its wallet balances dropping to approximately $69 million and its BMX token extending an 81.5% weekly decline. This event highlights the risks of centralized exchanges and the potential loss of user funds during wind-downs, eroding trust in the crypto exchange ecosystem and impacting users who may face long recovery times. The wallet balance drop from previous levels and the BMX token crash reflect market panic and reduced liquidity, with no clear timeline for full withdrawal processing or compensation.

rss · Cointelegraph · Jul 27, 12:34

Background: BitMart is a centralized cryptocurrency exchange that launched its native BMX token as an ERC-20 token to support its ecosystem. A wind-down means the exchange is ceasing operations, often leading to withdrawal halts and asset freezes. Similar events have occurred at other exchanges, causing significant user losses.

Storj files for Chapter 11 bankruptcy, explores equity for tokenholders ⭐️ 8.0/10

Decentralized storage provider Storj has filed for Chapter 11 bankruptcy protection, while exploring a court-approved mechanism to offer equity to STORJ tokenholders. The company assures that its network will continue operating during the process. This is a significant event for the decentralized storage sector, as one of its prominent players faces financial distress, potentially impacting user trust and industry confidence. The proposed equity path for tokenholders could set a precedent for how crypto token holders are treated in bankruptcy proceedings. Storj filed for Chapter 11, which allows it to restructure while continuing operations. The specific equity conversion mechanism for STORJ tokens requires court approval and details are not yet disclosed.

rss · Cointelegraph · Jul 27, 09:01

Background: Storj is a decentralized cloud storage platform that uses blockchain technology to allow users to rent out unused hard drive space. The STORJ token is used to pay for storage services and reward node operators. Filing for Chapter 11 bankruptcy does not necessarily mean liquidation; it is a reorganization process often used by companies to manage debt while continuing business.

Discussion: On the Storj community forum, some users had previously questioned the utility of the STORJ token, noting that aside from a 10% discount for paying with STORJ, there was limited demand driver. The bankruptcy filing and proposed equity conversion may reignite debates about token value and holder rights.

Binance Announces Delisting of 5 Altcoins ⭐️ 7.0/10

Binance has announced it will delist five altcoins from its platform, as reported by Koin Bülteni. The specific tokens and delisting date have not been disclosed in the initial report. Delistings by Binance, the world's largest cryptocurrency exchange, often lead to significant price drops and reduced liquidity for affected tokens. This move could impact investors holding these altcoins and signal Binance's ongoing review of listed assets. The report from Koin Bülteni lacks official confirmation from Binance and does not name the five altcoins or provide a timeline. Such announcements are typically accompanied by reasons like low trading volume or regulatory concerns.

google_news · Koin Bülteni · Jul 27, 18:14

Background: Cryptocurrency exchanges like Binance periodically delist tokens to maintain platform quality, comply with regulations, or protect users. Delisted tokens are removed from trading pairs, and users must withdraw them before a deadline or risk losing access.

go-ethereum v1.17.5: Maintenance Release with Pebble v2 and Amsterdam EIPs ⭐️ 7.0/10

The go-ethereum team released v1.17.5, a maintenance update that changes the default GOGC value from 20 to 50, adds Pebble v2 database support, and continues implementing Amsterdam hardfork EIPs. This release improves node performance by reducing garbage collection overhead and offers a more efficient database backend, while advancing the upcoming Amsterdam hardfork, which will significantly alter Ethereum's gas model and contract size limits. Node operators are strongly recommended to upgrade. Pebble v2 is used when bootstrapping a new database; existing Pebble v1 databases must be manually upgraded via 'geth db pebble-upgrade'. The GOGC change adjusts the garbage collector trigger threshold, trading higher memory usage for lower CPU overhead.

github · rjl493456442 · Jul 27, 16:17

Background: GOGC is a Go runtime environment variable that controls the garbage collector's target percentage: a higher value means the GC runs less frequently, reducing CPU time but increasing peak memory. Pebble is a key-value store written in Go, initially used in go-ethereum as an alternative to LevelDB. The Amsterdam hardfork is a planned Ethereum upgrade that includes multiple EIPs to improve gas accounting, increase contract size, and introduce block-level access lists, building on changes from earlier forks.

US citizen charged after GrapheneOS phone wipe at border ⭐️ 7.0/10

A US citizen is facing federal charges after using a duress PIN on his GrapheneOS phone, which automatically wiped the device during a border search at an airport. This marks the first known prosecution involving the use of a duress password at a US border. This case sets a legal precedent for the use of duress PINs and other security features that can automatically destroy data, raising important questions about the balance between individual privacy rights and government authority at borders. It could deter security-conscious users from utilizing such features. The duress PIN on GrapheneOS is designed to wipe the device when entered under coercion, but prosecutors argue the act of wiping constitutes obstruction of justice. The outcome could hinge on whether using a duress PIN is considered an intentional act to destroy evidence.

hackernews · eecc · Jul 27, 06:21 · Discussion

Background: GrapheneOS is a security-focused Android-based operating system that includes features like duress PINs to protect users' data. A duress PIN is a separate password that, when entered, triggers data deletion instead of unlocking the device. US border authorities have broad search powers, and individuals are generally expected to comply with device searches. This incident highlights the legal risks of using anti-forensic tools at borders.

Discussion: Commenters noted that while duress PINs are a useful security feature, users must recognize the potential legal consequences at borders. Some suggested using decoy OS volumes like VeraCrypt's hidden volumes as a less extreme alternative. Others argued that travelers should wipe their devices before crossing and restore from backup later.