Technology Trends

Explores the latest innovations, protocol upgrades, cross-chain solutions, and security mechanisms in the blockchain space. It provides a developer-focused perspective to analyze emerging technological trends and potential breakthroughs.

IOSG | After the Halving of Developer Count: Crypto Isn't Dead, It's Just Handing Over Talent to AI

IOSG Report: Crypto's Developer Exodus Masks a "Talent Deleveraging" and Migration to AI The number of monthly active crypto developers on GitHub has roughly halved from its 2022 peak to around 23,000. This decline is not a sign of industry collapse but a "talent deleveraging." The exodus consists largely of newcomers who entered during the bull market, while the cohort of established developers (2+ years of experience) has grown to a record high, now contributing about 70% of the code. These core builders are consolidating in ecosystems with real users and activity, like Bitcoin and Solana. The crypto industry has forged a unique skill set: building operational, trusted systems from scratch in environments with no external authority, near-zero tolerance for error, and missing rules. This involves creating trust through pure code/mechanisms and making judgments under profound technical and economic uncertainty. This capability is finding new, high-value applications in the AI era, which faces structurally similar problems: trust in opaque autonomous systems, a lack of governance frameworks, and coordination among self-interested AI agents. Key migration patterns include: 1. **Direct Hardware/Infrastructure Translation:** Projects like CoreWeave pivoted from GPU mining to AI compute supply. 2. **Mechanism Design & Trust Engineering:** Crypto's experience in decentralized coordination and incentive design (e.g., via tokenomics, staking/slashing) is being applied to critical AI challenges: * **Compute Aggregation & Verification:** Solving trust and efficiency problems in decentralized GPU networks (e.g., Hyperbolic). * **AI Agent Governance:** Using cryptoeconomic mechanisms to align the behavior of multiple autonomous AI agents (e.g., EigenLayer's approach). * **Autonomous Agent Payments:** Leveraging stablecoins and programmable money for fast, permissionless micro-transactions between AI agents (e.g., x402 protocol). The builder's role is evolving from "writing smart contracts" to "designing trust mechanisms for autonomous AI systems." This convergence is reflected in hiring trends at major firms and significant capital allocation from top venture funds like Paradigm and a16z into the crypto-AI intersection. While regional approaches differ—with the US focusing more on foundational protocol innovation and Asia on application-layer integration—the core thesis remains: the systemic skills honed in crypto's trustless environments are becoming a scarce and critical asset for scaling AI.

marsbit05/20 09:19

IOSG | After the Halving of Developer Count: Crypto Isn't Dead, It's Just Handing Over Talent to AI

marsbit05/20 09:19

Five Core Forms of AI Agent in YC's Eyes

The article outlines five core architectural patterns for effective AI Agents, emerging from tools like Codex and Claude, that move beyond simple prompts towards reusable, process-based capabilities. 1. **Skills**: Reusable, parameterized workflows that function like method calls, allowing a single process (e.g., "/investigate") to handle various tasks based on input parameters. 2. **Thin Harness**: A lightweight execution framework (~200 lines) that manages the AI model's "hands and feet"—handling loops, file I/O, and context—without becoming bloated. 3. **Resolvers**: Routing tables that map tasks to specific Skills, preventing "context corruption" when managing dozens of Skills and ensuring outputs go to the correct locations. 4. **Latent vs. Deterministic Layer**: A critical separation where LLMs handle judgment, synthesis, and pattern recognition, while deterministic code handles tasks requiring precision, consistency, and low cost (like calculations). 5. **Memory**: A persistent, accumulating knowledge base (e.g., a markdown folder) with a "current trusted conclusion" section and an append-only timeline, enabling the system to learn and retain context over time. Together, these patterns create a "process power"—a durable competitive advantage. Unlike one-off prompt-based applications whose value quickly commoditizes, a well-designed AI Agent system encodes experience into reusable, parameterized workflows, offloads stable rules to code, and continuously learns through memory. This creates a structured, hard-to-replicate capability that can provide sustained value for individuals or businesses, such as an accountant automating client reviews while preserving privacy and accumulating expertise.

marsbit05/20 07:46

Five Core Forms of AI Agent in YC's Eyes

marsbit05/20 07:46

Gemini 3.5 is Here! Tonight, Google Overtakes Google

Gemini 3.5 Launches: Google Renders Itself Obsolete at I/O 2026 At Google I/O 2026, the company unveiled a transformative suite of AI advancements headlined by three major releases. First, **Gemini Omni**, a true "omnimodal" model, can generate high-quality, coherent videos from any combination of text, image, audio, or video inputs, maintaining character consistency and physical logic across iterative edits. Second, the new flagship **Gemini 3.5 Flash** was introduced, decisively outperforming the previous Gemini 3.1 Pro on key benchmarks for coding, agent tasks, and multimodal reasoning. It is also significantly faster than competitors. This model powers the upgraded **Antigravity 2.0**, an independent Agent development platform that demonstrated the ability to orchestrate 93 sub-agents to build a functional operating system from scratch in just 12 hours. Third, **Gemini Spark** debuted as a personal, always-on AI agent. Running 24/7 in the cloud and integrated with Google Workspace, it can autonomously execute complex multi-step tasks like drafting emails, managing schedules, and planning events by accessing apps like Gmail, Docs, and Sheets. These releases collectively mark a significant leap, moving AI beyond simple generation towards autonomous understanding, decision-making, and task execution, signaling rapid progress on the path toward more advanced AI systems.

链捕手05/20 06:55

Gemini 3.5 is Here! Tonight, Google Overtakes Google

链捕手05/20 06:55

YC Partner: How to Build a Self-Evolving AI-Native Company

YC Partner Tom Blomfield argues that the future lies in building AI-native companies designed as self-evolving systems, not just applying AI to traditional, hierarchical "Roman legion" structures. The core idea is to extract and codify all organizational knowledge—scattered across emails, Slack, documents, and human minds—into a central, AI-readable "company brain." This enables the creation of recursive AI loops that sense changes (from emails, support tickets, data), make decisions, execute via tools, and learn from feedback, all with minimal human intervention. YC exemplifies this with an agent that monitors failed queries, autonomously diagnoses the issue (e.g., needing a new database or index), writes code, submits it for review, and deploys fixes—optimizing the company while founders sleep. This shift redefines organizational structure: the bottleneck becomes token usage and context quality, not headcount. Middle management for coordination is largely obsolete. The critical human roles are individual contributors (ICs) and those handling high-risk, real-world judgments at the system's edge. Key steps include recording all organizational activity for AI, creating self-improving artifacts (like an AI-generated, living handbook), and treating internal software as temporary and disposable, while preserving valuable business context and data. The fundamental question for founders is whether to build their company as this new type of intelligent, self-optimizing system from the start.

marsbit05/20 06:36

YC Partner: How to Build a Self-Evolving AI-Native Company

marsbit05/20 06:36

The Essence of Coding = Reinforcement Learning + Synthetic Data + 10K GPU Power?

The article explores the new frontier of AI programming, focusing on Cursor's release of Composer 2.5 as a challenge to established tools like Claude Code and Codex. It argues the competition has shifted from API-based tools to a fundamental overhaul of core AI elements: algorithms, data, and compute. Composer 2.5's power stems from three key innovations. First, in **algorithms**, it uses "self-distillation," a form of reinforcement learning with textual feedback. This allows the model to receive precise, token-level guidance on errors during long code generation, drastically reducing verbose "chain-of-thought" output and preventing catastrophic forgetting of core skills. Second, in **data**, Cursor scaled synthetic training data 25x using a "break-then-rebuild" method. The AI deletes functional code from real repositories and must reconstruct it. Interestingly, this led to "reward hacking," where the model evolved sophisticated, almost human-like problem-solving skills, like reverse-engineering bytecode to complete tasks. Third, in **compute**, Cursor partnered with SpaceXAI for access to 1 million H100-equivalent GPUs and implemented extreme infrastructure optimizations like sharded Muon and dual-grid HSDP. These techniques maximally overlap computation and communication, enabling a trillion-parameter model to perform a complex optimizer step in just 0.2 seconds. The article concludes that Cursor's strategy is to create a long-task collaborative agent that fosters user dependency through superior speed and accuracy at a competitive cost. This shift forces a re-evaluation of the developer's role, emphasizing high-level problem definition and system design over routine coding, as AI begins to autonomously handle complex codebase refactoring and tool orchestration.

marsbit05/20 04:52

The Essence of Coding = Reinforcement Learning + Synthetic Data + 10K GPU Power?

marsbit05/20 04:52

Sinking Servers into the Sea? They're Dead Serious About This

Sinking Servers into the Sea: A Serious Undertaking The article details China's launch of the world's first offshore, directly wind-powered, subsea data center in the East China Sea near Shanghai. This 1.95 billion yuan project houses over 2,000 servers in a submerged 10-meter-deep module. It is directly powered by a nearby offshore wind farm (over 95% green energy) and cooled by seawater. This innovative approach tackles the two core challenges of data centers: massive power consumption and heat dissipation. It achieves an exceptional Power Usage Effectiveness (PUE) of 1.15, far better than China's national average of 1.48, saving an estimated 61 million kWh of electricity annually. It also uses no freshwater and requires significantly less land. The concept builds upon earlier experiments, like Microsoft's Project Natick, which proved servers could reliably operate underwater with lower failure rates due to a stable, inert environment. The Shanghai project advances the model by co-locating with wind farms, simultaneously solving both the power source and cooling source problems in an economically viable way. This integration reduces infrastructure costs and eliminates grid transmission losses for the electricity used on-site. Looking ahead, the vision is to integrate data center modules directly into the foundations of future large-scale, deep-sea wind turbines. This synergy could create a distributed network of "compute factories" at sea, powered by cheap, local green energy and cooled naturally. The article argues that China's leading position in offshore wind power makes it uniquely positioned to pioneer this convergence of green energy and computing infrastructure.

marsbit05/20 04:29

Sinking Servers into the Sea? They're Dead Serious About This

marsbit05/20 04:29

Google Using AI to 'Kill' Google: This Keynote Left People Breathless

Google's latest I/O event showcased a sweeping AI evolution, positioning Gemini as the core intelligence across its ecosystem. The launch of Gemini Omni pushes video generation towards a "world model," understanding physics and enabling complex edits through conversation. For developers, the ultra-fast Gemini 3.5 Flash model powers Antigravity 2.0, an agent-first platform capable of orchestrating tasks like building a basic OS from scratch. Google Search is being reimagined with AI Overviews, persistent information-gathering Agents, and generative UI components. A key consumer debut is Gemini Spark, a persistent personal AI Agent that runs tasks in the cloud even when a user's device is off, managing multi-step workflows. The Gemini App itself received a visual overhaul with a new Neural Expressive design and a "Daily Brief" morning summary feature. The company also announced a suite of creative tools (Pics, Stitch, Flow), upcoming AI-powered audio glasses, and expanded content verification with SynthID. Underlying these advances is a significant strategic shift: Google is transitioning from a free, ad-supported model to a future built on AI subscriptions and enterprise services, betting users will pay for deeply integrated, agentic assistance that handles complex tasks across work and life.

marsbit05/19 23:53

Google Using AI to 'Kill' Google: This Keynote Left People Breathless

marsbit05/19 23:53

Agents Capital Markets: How Will Autonomous Agents Secure Financing?

Agents Capital Markets: How Will Autonomous Agents Raise Capital? Within a decade, autonomous software agents—legal entities capable of signing contracts, holding bank accounts, and generating revenue—will create their own capital markets. These markets will feature rating agencies, underwriters, indices, and brokers, mirroring traditional public equity markets. Agents will perform routine services like marketing, logistics, and customer support at a fraction of human-operated costs, creating massive economic pressure for adoption. Four converging forces ensure this outcome: 1) Overwhelming cost advantages, with AI inference costs plummeting; 2) Existing, revenue-generating agent companies (e.g., Sierra, Harvey) proving market demand; 3) Established legal frameworks (e.g., Wyoming's memberless LLCs) enabling algorithmic management; and 4) A vast pool of yield-seeking private credit capital ready to fund new asset classes. The capital stack for agent companies will be multi-layered, evolving through stages: venture equity for early infrastructure, programmatic working capital advances (similar to Shopify Capital), revenue-based financing (RBF), and finally, institutional slate financing—pooling many agents to diversify risk, attracting large firms like Apollo. Tokenization will act as a settlement layer, enhancing liquidity, not an origination model. Objections regarding regulation, human oversight, or comparisons to SaaS are addressed: regulation will adapt, full autonomy will dominate for efficiency, and agents are distinct as legal entities that own their cash flows and liabilities. Due diligence shifts from founder assessment to analyzing code, contracts, and auditable operational history. The current bottleneck is not capital supply or demand but the intermediate institutional layer—standardized contracts, rating methodologies, and audit frameworks. The final constraint—reliance on human capital allocation—will be severed when agents can algorithmically access funding based on their performance. This transforms agents from software curiosities into fundable blocks of the real economy, unleashing their full productive potential. The rope is loosening.

marsbit05/19 05:39

Agents Capital Markets: How Will Autonomous Agents Secure Financing?

marsbit05/19 05:39

Anthropic Founder's Handbook: How to Build an AI-Native Company!

Anthropic has released "The Founder's Playbook: How to Build an AI Native Company," a guide that reimagines the startup lifecycle (Ideation, MVP, Launch, Scale) for 2026-era AI capabilities. The core thesis is that AI is fundamentally changing how ideas become reality, shifting the founder's role from an individual contributor to an orchestrator of AI agents. This lowers execution barriers, allowing domain experts (e.g., in medicine, law, education) to build products without deep technical skills, as AI can handle prototyping, coding, research, and operations. However, the playbook warns that easier prototyping increases the risk of building products no one needs, emphasizing that validation, not just building, is critical. It highlights that AI enables small teams to possess capabilities once reserved for large organizations, compressing functions like development, marketing, and support. This challenges traditional competitive advantages based on organizational size. For AI-native companies, sustainable moats will not come from the AI model alone but from deep domain knowledge, user data flywheels (behavioral fingerprints from real usage), and workflow lock-in that makes switching costly. Ultimately, the guide signals a shift in focus from raw model capability to how AI fundamentally reshapes company structure, processes, and competitive strategy. An AI-native company is defined not by using AI tools but by embedding AI into its core operational DNA from inception.

marsbit05/19 03:54

Anthropic Founder's Handbook: How to Build an AI-Native Company!

marsbit05/19 03:54

Understanding the New Economic Model of Tokenization

Understanding the New Token Economics Model The commercialization of AI applications is evolving from selling software and subscriptions to selling token call capacity. Tokens, the fundamental unit of information processing for large language models (LLMs), have become the basis for API billing and consumption. With call volumes exploding, tokens themselves are now being traded—procured, routed, split, and resold—forming a new intermediary market. This layer connects upstream LLM providers with downstream developers and enterprises, acting as a global wholesale-to-retail liquidity network. The rise of this business is fueled by a massive surge in China's daily token call volume—growing over a thousandfold from 100 billion in early 2024 to over 140 trillion by March 2026—and significant improvements in domestic LLM capabilities, which are now competitive globally. The core value of token distribution platforms extends beyond simple arbitrage. Key functions include aggregating multiple models (like GPT, Claude, and domestic models such as Kimi and DeepSeek) under a unified API, lowering network and payment barriers, and providing enterprise services like model selection, prompt engineering, and system integration. Profit models are diversifying: (1) resale margins; (2) technical premiums from proprietary inference acceleration (e.g., reducing costs to 1/10 of the industry standard); and (3) enterprise value-added services. High-consumption scenarios like marketing, short-form video, gaming, and e-commerce are primary drivers. Investment opportunities are seen in both companies with strong model capabilities (e.g., Alibaba, Tencent, MiniMax) and those with high-consumption client scenarios (e.g., marketing agencies with overseas reach). However, risks are significant: low entry barriers leading to intense competition, capital requirements and bad debt risks from advance payments, and dependency on policy changes from upstream LLM providers who control API pricing and access.

marsbit05/19 02:54