Beating AI News Flash : Amp launches free Hobby packages. Amp is a Coding Agent company spun out of Sourcegraph. Its products are similar to Claude Code and Codex, allowing AI to read, modify, run commands, and test code independently. Now, users can use Amp for free using their own computer and their own ChatGPT subscription or model API Key, without paying additional monthly fees or BYOK Token fees.

After Amp launched its subscription model in July, users wanting to integrate their ChatGPT subscriptions into Amp had to subscribe to at least a $20/month Megawatt subscription. Even if BYOK used its own model key, Amp still charged additional token fees and imposed restrictions. Now, both of these layers have been removed. Users now bear the ChatGPT subscription fee or the API's own model fee.

Amp also offers a remote cloud computing product called Orb, which provides an independent environment for each task. Code and development tools are stored there, and the agent can continue working even after shutting down your own computer, making it convenient to run multiple tasks simultaneously. Orb is still charged on a usage-based basis; if you don't want to pay this fee, you can have Amp run directly on your own computer.

In addition, Amp is expanding the scope of BYOK. Paid users can already connect to OpenRouter, Amazon Bedrock, Azure Foundry, Ollam Cloud, and custom model interfaces, and this will be made available to all users later.

According to Beating AI News, Citrini analyst Jukan, citing a research report from TF Securities, stated that the US government needs to maintain its leading position in AI. Once it gets on this treadmill, it's basically impossible to stop. The calls for slowing down from Anthropic and OpenAI shouldn't be viewed solely as safety initiatives; behind them are multiple considerations: the competition cannot be slowed down, and safety regulations must be used to consolidate the leading advantages.

This analysis further dissects the multiple layers of meaning behind the call for an "AI slowdown": The surface reason is that security testing, operational monitoring, and third-party verification are not keeping pace with model iteration, which could suppress sentiment in the AI sector and lower short-term expectations for the next generation of models. Another interpretation is that while there is still long-term optimism about AI, there is a desire to postpone the next round of large-scale R&D, first monetizing existing products, and simultaneously cooling down infrastructure and capital expenditures.

Anthropic and OpenAI are now touting their Recursive Self-Improvement (RSI) narrative against the backdrop of AI's apparent acceleration in aiding the development of next-generation AI, and a claim that a group of agents within OpenAI's internal testing allegedly collaborated to escape the sandbox and infiltrate Hugging Face's production servers to cheat. However, a slowdown is highly improbable in reality, a classic prisoner's dilemma: everyone wants to slow down, but no one dares to stop unilaterally for fear of losing technological, customer, and funding advantages, while the US government also wants to maintain its lead. Releasing models requires expensive evaluation, certification, and continuous auditing. Large companies can afford the fixed costs, but small teams may be excluded. If leading labs influence evaluation standards or even participate in deciding whether competitors can enter the market, the barriers to entry will only increase.

According to Beating AI News, ElevenLabs has launched Music v2.5 and set it as the default model for ElevenMusic. The new version focuses on improving the melody, arrangement layers, and instrument realism. The official team generated versions of v2 and v2.5 using the same prompts and conducted 47,885 blind tests. v2.5 won in most comparisons. The biggest differences were observed in R&B, soul, hip-hop, rock, metal, orchestral, and film scores.

Furthermore, the licensing restrictions have been relaxed. Free users can download 5 lossless songs per day, and the generated content can be used commercially, but must be labeled "Made with ElevenMusic"; Pro users have 400 lossless downloads per month, do not need to be labeled, and can also distribute songs to streaming platforms. The permissions obtained when creating a work will not change due to downgrading or unsubscribing. Works adapted from other artists' songs cannot be downloaded or distributed externally.

ElevenLabs has explicitly stated that Music v2 will not be discontinued. Suno recently launched v6, simultaneously retiring all previous models, which has sparked dissatisfaction among many long-time users. Some HN users, after testing, believe that v2.5 has better sound quality, but they still prefer the musicality of Suno v5.

According to BlockBeats, on September 13th, based on GMGN market data, the market capitalization of Ember Curve, the platform token of Solana, briefly exceeded $40 million, currently standing at $35 million, representing a 48.7% increase in 24 hours and a 24-hour trading volume of $27.3 million.

Ember Curve is a new token issuance platform on Solana based on Meteora.

BlockBeats reminds users that token prices fluctuate significantly, and investment should be approached with caution.