mod

Pump.fun announced the launch of the "Holder Reward" mechanism and the cancellation of the Cashback model. From now on, when token creators issue new coins, they can choose between the standard "Creator Fee" model and the "Holder Reward" model.In the Holder Reward model, the relevant fees will be transferred to the Pump.fun distribution wallet, and the platform will automatically conduct multiple proportional distributions every hour. Users with a holding value exceeding 20 USD will receive rewards, with more holdings resulting in more rewards. Rewards are paid in the quoted token of the trading pair; for example, tokens paired with SOL will distribute rewards in SOL.Existing Cashback and Creator Fee model tokens can apply to be converted into Holder Reward tokens, and the conversion cannot be reversed. The SOL and USDC trading pairs will continue to use a tiered fee rate that decreases as market value increases; custom trading pairs can set a fixed rate between 0.01% and 3%, which cannot be modified once determined.

The Solana meme coin launch platform Pump.fun announced two updates: the introduction of holder rewards and the removal of the cashback model. Holder reward tokens can earn rewards simply by holding them, with higher limits for longer holding periods, aimed at community development.The cashback model is no longer an option for launches. Creators can choose between creator fees or holder reward tokens. Existing tokens can apply for an irreversible switch, submitted by the team or community, with review criteria based on objective information beneficial to the token and community. Custom trading pairs must set a fixed fee rate of 0.01% to 3%, while SOL/USDC will continue to use the market cap tiered fee rate.Fees go into the Pump.fun distribution wallet, automatically distributed multiple times per hour based on holding proportions, with participation allowed for holdings over $20 in value. Protocol fees are the same across all types, and rewards are distributed in quoted tokens.

Anthropic CEO Dario Amodei called for a slowdown in AI model development and the introduction of third-party evaluation agencies. OpenAI CEO Sam Altman stated that he would adopt the suggestions, and Elon Musk also expressed support.Despite the warnings from the three leaders about AI risks, there are few precedents in the industry for coordinating a slowdown in development speed.

On September 12, B.AI announced that a new round of rights for popular models officially took effect: the high-concurrency main models DeepSeek-V4.1-Flash and GLM-5.3-Flash are now available for a special 90% discount, allowing users to access them for only 10% of the original price, catering to high-throughput scenario demands with exceptional cost-effectiveness. At the same time, Qwen3.8-Flash, Hy3, and MiMo-V2.5 continue to offer full free access, providing developers with richer zero-cost options.This adjustment in rights aims to meet high-throughput demands with exceptional cost-effectiveness, lowering the exploration threshold for developers through a multi-model matrix. B.AI continues to support developers in building low-cost AI infrastructure with a multi-model matrix and inclusive pricing, bidding farewell to high computing costs and accelerating innovation implementation. Start now: https://chat.b.ai/chat

Brian Armstrong posted on the X platform that Coinbase Wallet has launched Pulse mode, and users can download the app to experience it.

According to GMGN market data, the on-chain token issuance platform Commodity Market Exchange (CME) has a market capitalization exceeding 15 million USD, currently reported at 15.85 million USD, with a daily increase of 14 times.Public information shows that the platform supports token issuance paired with tokenized commodities as the underlying assets.

According to CNBC, Y Combinator CEO Garry Tan stated at the YC annual Demo Day that regarding the distillation of OpenAI and Anthropic models by Chinese companies (including DeepSeek, Moonshot AI, MiniMax, etc.), he chose to "do nothing" and suggested that regulators focus instead on maintaining a commercial balance between open-weight models and cutting-edge models.Tan also holds a reserved attitude towards the doomsday theory of AI, believing that cybersecurity is currently the most urgent risk, while the employment impact and economic transformation caused by AI will take decades to truly permeate society.

The B.AI platform announces the official launch of the DeepSeek native multimodal MoE large model DeepSeek-V4.1-Flash. This model uses a backbone with 552 billion parameters and a Causal Encoder-Decoder architecture, with asymmetric computational scales for input and output.The model supports a context of 1 million tokens and 384K output, aimed at long-cycle programming agents, multimodal workflows, and high-concurrency agent systems. Currently, both Web Chat and API access have been opened.

According to the DeepSeek Harness team, the new version DeepSeek Harness v0.1.5 has been deeply debugged with the DeepSeek V4.1 Flash model, covering standard mode, Programmatic Tool Calling (PTC) mode, and minimalist mode, and supports updating system prompts while retaining the existing KV Cache.The web version adds support for uploading various file types such as images and PDFs, as well as a sidebar file tree preview, allowing direct viewing of Markdown, HTML, PDF, code, and images. The interface provides standardized extension entry points for plugins on both the left and right sides, supporting global entry, multiple tabs on the right, and column display.Long session loading and storage performance have been optimized, and bidirectional communication between the parent Agent and child Agents is more flexible, with the experimental plugin "Agent Teams" enabling multi-Agent team collaboration and task splitting.

DeepSeek V4.1 Flash has started launching the app, merging the original three modes: "Quick," "Expert," and "Image Recognition" into a single entry point. Users no longer need to manually select a mode; one model handles daily conversations, image understanding, and complex questions simultaneously.V4.1 Flash natively supports multimodal capabilities. DeepSeek previously stated that the new model has comprehensively surpassed V4 Pro in performance, cost, speed, and total time metrics, and will take over all V4 Pro requests before the launch of V4.1 Pro.

Pantera Capital Junior Partner Jay Yu published an article titled "The Rise of Compute Markets," stating that computing power and data center spending have become a trillion-dollar category, but GPU procurement is still mainly done through group chats, over-the-counter brokers, and bilateral agreements. He noted that the financialization of computing power is still in its early stages, constrained by SKU, time, and location, but could become a commodity asset similar to electricity or oil in the next 5 to 10 years. Jay Yu believes that the computing power market may form a "hardware-supplier-cluster" structure similar to the electricity "grid-operator-node," with new cloud vendors being structural shorts on GPUs, while on-demand platforms and application layers are longs.He mentioned that for every $100 spent on inference at the application layer, about $45 flows to the on-demand layer, about $50 flows to the new cloud or GPU layer, and about $5 flows to routing layers like OpenRouter. NVIDIA can be seen as the "central bank" of computing power, having announced support for up to 25% residual value. He also stated that physical delivery has a more enduring moat, and indices, cash-settled exchanges, and financing tools will subsequently emerge, with participants including SF Compute, Vast AI, Runpod, Compute Exchange, and others. The computing power market does not have mandatory price transparency requirements, and basis risk may be greater.

According to a report by the Financial Associated Press, the Chief Financial Officer of OpenAI stated that they are considering reforming the charging model, testing a new pricing model based on performance, gradually moving away from the usage-based billing method, and competing with lower-priced open-source models.

According to a report by Sina Finance, Anthropic refused to submit its latest model Claude Mythos 5.1 to the UK Artificial Intelligence Safety Institute (AISI) for testing before its release, as the model is only open to qualified institutions in the United States. This move has raised alarms within the UK Whitehall and AISI, with UK national security officials concerned that this could set a precedent, making it difficult for AISI to access the latest cutting-edge models in the future.Analysts pointed out that this action may be related to the increasingly stringent export control policies of the US government on the AI industry. The US government had previously implemented export controls on Anthropic's models Mythos and Fable, prohibiting foreign nationals from using them. Anthropic declined to comment on this, only stating that it is coordinating with the US government to expand access to the model for more domestic and international partners as soon as possible.

According to official news, the Circle Cross-Chain Transfer Protocol (CCTP) now supports a fast transfer prepayment fee feature. Previously, the protocol fee would be directly deducted from the USDC transfer amount on the target chain; with this update, developers can pre-quote and collect fees using the native Gas token or USDC from the source chain before executing the transfer, thereby simplifying the fee handling for cross-chain transfers and ensuring that users receive the expected amount of USDC.It is reported that the prepayment fee mechanism mainly improves the predictability of fees in cross-chain applications. Users do not have to bear the reduction in the actual amount received due to additional deductions on the target chain; at the same time, the CCTP Quote API can uniformly calculate fees for different supported chains and integrate Fast Transfer and Forwarding fees into a single quote, so developers do not need to build systems to separately calculate multiple protocol fees. In addition, fees can be paid directly using the native token of the source chain, which will not reduce or affect the actual transferred USDC balance.Currently, the prepayment fee feature supports USDC transfers on all EVM chains covered by CCTP, but it does not support transfers initiated from Solana; transfers to Solana are still supported. Developers can obtain specific fee quotes through the Quote API and integrate this feature, and relevant developers can refer to the CCTP official documentation to build a more predictable cross-chain USDC transfer process.

According to a post by X platform user "Yongchutafi" (@xhyctf) on September 8, DeepSeek may launch a new model V4.1 Flash.

According to Bloomberg, ByteDance is preparing to launch an artificial intelligence model for real-time spatial video generation, with applications covering fields such as robotics and autonomous systems, and will compete with companies like Meta and Alphabet.Insiders say that ByteDance founder Zhang Yiming is personally overseeing the development of the model, coordinating multiple business departments within the company, and investing in artificial intelligence resources and computing power support. The model may be released as early as next month, but the specific timing is still subject to change.

Goldman Sachs significantly raised its industry forecast in the latest global optical module report, expecting the global optical module market size to reach approximately $67.7 billion, $131.4 billion, and $148.5 billion in 2026, 2027, and 2028 respectively, an increase of 33%, 81%, and 115% compared to previous forecasts.Goldman Sachs believes that this round of upward revision is not solely due to the increase in the number of AI servers; more importantly, the architecture of AI servers is continuously migrating towards high-speed networks, coupled with the increase in the number of optical modules required for a single GPU or ASIC chip, driving rapid growth in demand for 800G, 1.6T, and even 3.2T.

The Sun Wukong ecosystem will hold a roundtable discussion at X Space on September 8th at 8 PM, with the theme "GPT-6, Claude, Grok taking turns to upgrade, the AI model war has entered the second half."This event will focus on the shift in industry competition amid the convergence of leading large model capabilities, and will delve into the evolution of true industry barriers from model parameters to underlying computing power scheduling, multi-model API routing efficiency, and the infrastructure supporting Agent automated execution.