# Open-Source Related Articles

HTX News Center provides the latest articles and in-depth analysis on "Open-Source", covering market trends, project updates, tech developments, and regulatory policies in the crypto industry.

Just now, DeepSeek V4 updates with DSpark, improving inference speed by 80%

DeepSeek has updated its DeepSeek V4 model with the DSpark speculative decoding framework, achieving a significant 60-85% speedup in generation for Flash models and 57-78% for Pro models while maintaining the same overall throughput. This engineering-focused update, rather than a core architectural change, introduces DSpark to address latency and throughput bottlenecks in high-concurrency production environments. DSpark combines high-throughput parallel generation with adaptive load-aware verification. Its key innovations include a semi-autoregressive generation architecture to model dependencies within token blocks and a hardware-aware confidence-scheduled verification system. This system uses a confidence head to predict token acceptance probabilities, allowing it to dynamically optimize verification length per request and allocate compute only to tokens with the highest expected payoff. The asynchronous scheduler is designed for real-world deployment, ensuring zero-overhead scheduling and continuous CUDA graph replay while preserving the target model's output distribution. In tests across mathematical reasoning, code generation, and daily dialogue, DSpark outperformed state-of-the-art models like Eagle3 and DFlash, increasing average acceptance length by 26.7%-30.9% and 16.3%-18.4% respectively on Qwen3 target models. DeepSeek also open-sourced DeepSpec, a full-stack codebase for training and evaluating speculative decoding draft models, providing a standardized toolkit that includes data preparation tools, model implementations, training code, and evaluation scripts.

marsbit06/27 08:52

Just now, DeepSeek V4 updates with DSpark, improving inference speed by 80%

marsbit06/27 08:52

28,000-Star Hit Creator Unexpectedly Fired by Google, Silicon Valley Big Shots Quickly Vie for the Talent

A nearly 7-year Google veteran, Justin Poehnelt, was fired shortly after his open-source project 'gws' (a Google Workspace CLI built in Rust) went viral, amassing over 28,000 GitHub stars. The tool, designed for both humans and AI agents, allowed terminal access to all major Workspace services like Gmail and Drive by dynamically generating commands from Google's own APIs. The dismissal came just two days after Google announced its "official" Workspace CLI at its Cloud Next 2026 conference, leading to widespread criticism that the company appropriated his work. Poehnelt believes the real reason was internal fear that his project demonstrated how AI agents could fundamentally disrupt and simplify interaction with Google's core Workspace products, potentially threatening its established business model. Following his public disclosure of the events, which garnered millions of views, tech leaders like OpenClaw's Peter Steinberger and CEOs from OpenRouter and Vercel swiftly offered him jobs, highlighting the perceived misstep by Google in letting a top-performing innovator go.

marsbit06/26 00:19

28,000-Star Hit Creator Unexpectedly Fired by Google, Silicon Valley Big Shots Quickly Vie for the Talent

marsbit06/26 00:19

AI Stocks Plunge Violently, Is It GLM's DeepSeek Moment?

On Tuesday, AI-related stocks faced their most severe sell-off of the year, triggered by a sharp decline in South Korean equities. The plunge, led by key AI supply chain players Samsung and SK Hynix, spread to US markets, heavily impacting semiconductor, memory, and storage stocks. Analysts linked the downturn to the recent release of China's powerful open-source model, GLM-5.2, drawing parallels to the market shock caused by DeepSeek's release earlier in 2025. This event reignited investor concerns that cheaper, competitive AI models could undermine the economic rationale for the massive data center investments by major US tech firms. The sell-off was seen as a repricing of the AI trade rather than a rejection of AI demand itself. Market focus shifted from whether AI will grow to whether the current valuations justify the enormous capital expenditures. While some strategists view this as a necessary correction after excessive gains, questions remain about the sustainability of financing for AI infrastructure, particularly as companies increasingly rely on debt. All eyes are now on upcoming earnings reports, like Micron's, to gauge the hardware sector's health. The core debate has become about the cost of growth and which companies can successfully convert their massive investments into sustainable cash flow.

marsbit06/24 02:21

AI Stocks Plunge Violently, Is It GLM's DeepSeek Moment?

marsbit06/24 02:21

As Channel Dividends Fade, What Can DeFi Protocols Rely On to Resist Giant Harvesting?

The article discusses how major tech companies like Coinbase, Stripe, and Kraken are increasingly acquiring or building their own underlying infrastructure to capture profits, moving away from relying on external DeFi protocols. It uses the example of Coinbase's Base blockchain, which generates sequencing fees for Coinbase from protocols like Morpho that deploy on it. Similarly, Stripe acquired Bridge to capture stablecoin yield instead of paying it to Circle, and Kraken bought NinjaTrader for its derivatives licenses. The core argument is that when channel-based advantages (like user traffic) fade, control over the profit-generating infrastructure becomes critical. While this could lead to a future dominated by a few integrated giants, the article suggests a more balanced outcome is possible. Protocols that have already achieved deep, multi-chain integration and embedded themselves into corporate backends create high switching costs, making them difficult to replace. Examples include Morpho (despite reliance on Base, it's widely deployed) and Uniswap. The piece concludes that the race between institutional expansion and open-source protocol proliferation will ultimately shape the industry's structure.

marsbit06/22 10:22

As Channel Dividends Fade, What Can DeFi Protocols Rely On to Resist Giant Harvesting?

marsbit06/22 10:22

Channel Dividends Are Dwindling, What Can DeFi Protocols Rely on to Resist Giants' Harvesting?

As the era of channel-driven growth ends, major tech giants like Coinbase, Stripe, and Kraken are vertically integrating by acquiring or building core infrastructure, capturing profits from underlying protocols. For instance, Coinbase leverages its Base blockchain to earn sequencing fees from protocols like Morpho, while Stripe acquired Bridge to internalize stablecoin yields, and Kraken bought NinjaTrader for derivatives licenses. This trend creates a power imbalance: entities controlling distribution channels can extract value from open-source DeFi protocols. However, protocols are developing a defense through multi-chain expansion. Examples like Morpho (with significant TVL on Ethereum and Base) and Uniswap (dominant across many chains) show that by deploying across multiple blockchains, protocols reduce dependency on any single chain operator, making them costly and risky to replace. The article argues that the future landscape will be shaped by the race between institutional vertical integration and protocol horizontal, multi-chain scaling. Protocols that embed themselves deeply across ecosystems and become indispensable due to high switching costs are most likely to survive and thrive amidst this consolidation.

Foresight News06/22 08:12

Channel Dividends Are Dwindling, What Can DeFi Protocols Rely on to Resist Giants' Harvesting?

Foresight News06/22 08:12

OpenAI's "Most Open" Move: Codex No Longer Exclusively Favors GPT

OpenAI has significantly opened up its Codex programming agent by introducing a "model provider" configuration layer that allows users to connect it with various open-source models, not just its proprietary GPT. Through a configuration file or a simple `--oss` command-line flag, Codex can now route requests to local services like Ollama or LM Studio, or to third-party APIs such as Mistral or DeepSeek. This move is seen as one of OpenAI's most "open" steps, potentially lowering costs and enhancing privacy for developers who can run code generation offline. However, integration isn't seamless for all models. Codex primarily uses OpenAI's newer Responses API, while many open-source models rely on the older Chat Completions interface. This creates compatibility issues, especially for advanced features like function calling. The developer community is already building "routing" or adapter layers (e.g., CC Switch, LiteLLM) to translate between these protocols, enabling hybrid setups where GPT handles planning and open-source models handle execution. Analysts interpret this as a strategic shift for OpenAI: from competing solely on model superiority to controlling the platform and interface standards. By making Codex a flexible, pluggable entry point for AI-assisted programming, OpenAI aims to become the central hub in the developer toolchain ecosystem, even as users gain the freedom to switch underlying models.

marsbit06/22 00:24

OpenAI's "Most Open" Move: Codex No Longer Exclusively Favors GPT

marsbit06/22 00:24

To 'Own' or 'Rent' Intelligence? The New Question for AI Startups

The shutdown of Mythos has prompted a fundamental question for AI startups: is your intelligence "rented" or "owned"? While using frontier model APIs allows for rapid prototyping, it cedes control over your core capabilities to external providers, whose decisions on pricing, rules, or even shutdowns can jeopardize your business. The critical issue isn't just cost, but control and ownership. The lesson isn't to abandon frontier models, but to build upon them. "Owning" intelligence means starting with a powerful open-source model and systematically shaping it with your unique assets: your proprietary data, workflows, domain expertise, edge cases, and evaluation standards. Over time, this creates a company-specific asset that reflects your real work. The future of AI isn't a single, dominant model. There will be multiple "frontiers": general-purpose frontier models, company-specific models fine-tuned on proprietary knowledge, specialized vertical models, and routing systems that orchestrate multiple models. The winning companies will be those that transform intelligence into their own unique, controlled asset, ensuring no one can pull the floor from under their product.

marsbit06/19 02:08

To 'Own' or 'Rent' Intelligence? The New Question for AI Startups

marsbit06/19 02:08

From Ethereum to AI's 'CROPS': What Exactly Is This 'Slow Variable' That Vitalik Has Repeatedly Emphasized?

Recently, Vitalik Buterin has frequently emphasized the concept of "CROPS," first outlined in the Ethereum Foundation's March mandate as core principles guiding its focus: Censorship Resistance, Capture Resistance, Open Source, Privacy, and Security. CROPS represents Ethereum's commitment to providing foundational capabilities for user sovereignty—enabling asset ownership, identity expression, and coordination without reliance on centralized platforms or surrendering ultimate control. This framework is gaining new urgency with the rise of AI, particularly AI agents managing digital assets and automating transactions. While AI offers convenience, it risks centralizing user data, intent, and control if dependent on opaque, centralized services. Vitalik argues for "CROPS AI"—AI that is open, privacy-preserving, secure, and capable of local execution to maintain user agency. He highlights convergence between "CROPS Ethereum access layers" and "CROPS AI," such as using zero-knowledge proofs for private remote LLM calls and Ethereum RPC reads, ensuring users can access services without exposing sensitive information. Ultimately, CROPS is not just an abstract ideal but a practical guide for Ethereum's development and AI integration. It addresses the critical long-term question: as digital systems grow more powerful, how can users retain control over their privacy, assets, and autonomy? In an AI-driven era, these principles may define Ethereum's enduring value—prioritizing verifiable, secure, and user-centric design over short-term optimizations like speed and cost alone.

marsbit06/06 08:07

From Ethereum to AI's 'CROPS': What Exactly Is This 'Slow Variable' That Vitalik Has Repeatedly Emphasized?

marsbit06/06 08:07

Biology's Paradigm Shift: Zuckerberg's New Open-Source Model Completely Overturns Google's AlphaFold Throne

The AlphaFold era faces a major challenge. A new open-source AI model, ESMFold2, from Meta CEO Mark Zuckerberg's Biohub, has been released alongside a massive database of 11 billion predicted protein structures—surpassing the AlphaFold database by 8 billion entries. Published in Nature, the model is reported to outperform AlphaFold3 in key areas, particularly in predicting protein complexes. Crucially, it is fully open-source with no commercial restrictions. ESMFold2 takes a different technical approach, building on a protein language model trained on billions of sequences, including microbial data from diverse environments like soil and ocean—areas less covered by AlphaFold. The team validated its utility by designing and successfully synthesizing novel functional proteins in the lab. The decision to open-source everything is seen as a strategic move, similar to Meta's approach with its Llama models, aiming to build an ecosystem and accelerate global research. While scientists welcome the resource, some urge caution, noting the need for independent validation of predictions and questioning its performance on entirely novel protein folds. The development signals intensified competition in protein AI, rapidly evolving much like the large language model field, and represents a significant step forward in using AI to decode and engineer the machinery of life.

marsbit05/29 12:31

Biology's Paradigm Shift: Zuckerberg's New Open-Source Model Completely Overturns Google's AlphaFold Throne

marsbit05/29 12:31

DeepSeek's $10 Trillion Path: Leveraging Open Source to Pivot a Trillion-Dollar Hardware Ecosystem

This article proposes a strategic analysis of DeepSeek, arguing its goal is not short-term application-layer monetization but a long-term, foundational play to reshape the AI hardware ecosystem. It posits that DeepSeek's series of architectural innovations—MoE, MLA, DSA, CSA, Engram, TileLang—are fundamentally designed to drastically reduce the computational and memory (especially HBM) requirements for training and serving state-of-the-art AI models. By making AI feasible on hardware with lower performance (e.g., using NAND/SSD for KV Cache, LPDDR for weights and Engram storage), DeepSeek aims to foster and benefit from a viable alternative AI hardware supply chain, particularly in China. This strategy could unlock a trillion-dollar valuation for DeepSeek by enabling and owning a stake in a new, massive hardware ecosystem, rather than competing directly on subscriptions or multi-modal features.

marsbit05/25 13:13

DeepSeek's $10 Trillion Path: Leveraging Open Source to Pivot a Trillion-Dollar Hardware Ecosystem

marsbit05/25 13:13

活动图片

Ethereumright-arrow

Bitcoinright-arrow