# Workflow Related Articles

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

10,000 Scientists Get 1 Year of Free Access: OpenAI Brings the Scientific Research Pipeline into ChatGPT

OpenAI has launched the "ChatGPT for Academic Researchers" program, offering free one-year access to its flagship models for 100,000 university researchers globally, with 10,000 spots available this summer. Selected institutions include prestigious centers like ENS Paris and the IAS at Princeton. The initiative provides an integrated research workspace within ChatGPT, bundling tools like ChatGPT, ChatGPT Work, and Codex, along with expanded Deep Research capabilities, higher usage limits, and specialized tools for life sciences. The suite connects to platforms like Zotero and GitHub, aiming to streamline the entire research workflow from literature review and coding to data analysis and manuscript drafting. OpenAI notes that about 1.3 million people already use ChatGPT weekly for advanced science and math. The program targets building long-term user dependency by embedding these tools into daily research habits. However, access comes with limitations: it does not include API credits or model weights, and eligibility is restricted to verified academic researchers from supported countries. This approach contrasts with Anthropic's "AI for Science" program, which offers API credits but not an integrated workspace. Both companies emphasize preventing misuse by withholding model weights, a point of contention for AI researchers seeking transparency. The core strategy remains clear: provide a powerful, integrated environment to foster user reliance ahead of the post-free period.

marsbit10h ago

10,000 Scientists Get 1 Year of Free Access: OpenAI Brings the Scientific Research Pipeline into ChatGPT

marsbit10h ago

Claude Designer Lags Behind Engineers, Fires Back by Creating a Million-User Tool

The article tells the story of Nate Parrott, a designer at Anthropic who created Claude Design as a side project to keep pace with his engineering teammates. When Anthropic released Opus 4.5 in November 2025, the two engineers on Parrott's Claude Code team significantly increased their output using the new AI capabilities. Parrott, the sole designer, found himself struggling to match their speed, becoming a bottleneck in the workflow. To catch up, he began experimenting in his spare time. He initially tried prompting Claude to generate designs from text descriptions and screenshots, with limited success. His breakthrough came when he shifted focus from asking Claude to "design" to asking it to generate HTML. He realized HTML could be a rich visual canvas for creating everything from slides and interactive prototypes to full web pages. He built a simple interface with a chat panel on the left and a live HTML preview on the right. The key to making the output useful was incorporating Anthropic's brand system—fonts, colors, assets, and design principles—into the prompts. This ensured generated designs were immediately on-brand. He shared an internal prototype with his team, and other product designers quickly adopted it for creating clickable prototypes, a task traditionally requiring manually drawing every state. The project's "official" turning point came during an Anthropic Labs offsite, where Parrott noticed many attendees were using his tool to build presentation slides on the fly, sometimes right before their turn to speak. This organic adoption convinced the Labs team to formally staff the project, turning the side project into a real product. Claude Design is positioned as a "pre-production" tool for visual communication and exploration—handling slides, landing pages, PDFs, emails, and social media graphics. It integrates with tools like Canva, Adobe, and Vercel. Its core value is accelerating the early stages of design: exploring directions, building consensus, and establishing systems. For actual production code, Anthropic still recommends Claude Code. The story highlights how AI disrupts workflows unevenly and how individuals can respond by building new tools to create their own advantages. Parrott's tool, born from necessity, eventually gained over a million users in its first week.

marsbit2 days ago 00:41

Claude Designer Lags Behind Engineers, Fires Back by Creating a Million-User Tool

marsbit2 days ago 00:41

ChatGPT Voice Enters the Desktop Arena: You Talk, a Team of AIs Get to Work

Recently, AI pioneer Andrej Karpathy shared his preferred method for interacting with AI: using voice input to convey complex, unstructured thoughts. Instead of carefully typing out requests, he simply speaks freely for minutes at a time. He finds that AI excels at untangling these "stream-of-consciousness" monologues, returning clearer, more organized outputs than the original spoken input. This approach, he argues, improves "mind-merging" with the model by providing high-bandwidth context. Following this trend, OpenAI has integrated advanced voice capabilities into the desktop version of ChatGPT for Work and Codex scenarios, available globally for macOS and Windows. Powered by the new GPT-Live model, the feature allows real-time, interruptible conversation—users can speak to start tasks, check progress, manage multiple AI agents simultaneously, and change directions mid-task. The system can leverage project context, connected documents, calendars, and communication tools to continue unfinished work. Industry leaders like Elon Musk and Sam Altman have also emphasized the shift toward voice, noting that typing is a bottleneck for conveying complex intent to increasingly capable AI agents. Speech offers a much higher throughput, allowing users to dump extensive context quickly, even if it's messy, letting the AI handle the structuring. Beyond mere input, OpenAI's desktop voice feature enables project management through speech. Users can prioritize tasks, coordinate multiple agents, and receive audio or visual notifications on task completion or blockers. This positions ChatGPT not just as a conversational tool but as an "AI work operating system"—a central hub where users act as managers orchestrating a team of AI assistants to handle tasks across files, codebases, and software on the desktop, the primary workspace for complex work. The core insight is that by removing the input bottleneck, voice interaction allows humans to focus mental energy on higher-level decision-making, while delegating execution to an ever-present team of AI agents.

marsbit07/27 10:42

ChatGPT Voice Enters the Desktop Arena: You Talk, a Team of AIs Get to Work

marsbit07/27 10:42

Claude Doesn't Submit Code Directly After Writing It: Runs 4 Skills for Self-Check, Fixes Issues, Then Comes Back to You

Claude No Longer Submits Code Directly: 4 Self-Check Skills to Run Before Coming Back to You AI already writes code, but the burden of reviewing it still falls on you. To address this, Anthropic has built a "verification loop" into Claude Code. After writing code, Claude now runs four self-check skills before delivering the work: * `/code-review`: Finds potential bugs and provides review feedback. * `/simplify`: Cleans up the diff, removing redundant or over-complex implementations to reduce future maintenance costs. * `/verify`: Performs end-to-end validation, actually running the application to confirm the feature works, not just appears to. * `/design`: Used only for UI changes; cross-references the implementation against the project's DESIGN.md file. This loop extends the AI agent's workflow from "gather context → execute" to "gather context → execute → auto-verify → fix → re-verify." It tackles the new bottleneck in AI-assisted development: the speed of verifying code now outpaces human review. These skills are built on Claude Code's existing verification foundation (like running apps and using linters). Teams can create their own custom verification skills by documenting their repetitive manual checks in plain language as Markdown files. Verification can be triggered at four levels: manually (Standalone), embedded in a task, chained with other skills, or automatically on every PR (On every PR). The shift signifies that competition in AI programming is moving from code generation to robust verification and self-correction. Well-built verification loops allow AI agents to run longer and more autonomously with less human supervision. Skills, which encapsulate team knowledge and workflows, are becoming a cross-vendor standard, meaning a team's efficiency gap will depend less on the AI model and more on their investment in these automated workflows and verification mechanisms.

marsbit07/27 10:23

Claude Doesn't Submit Code Directly After Writing It: Runs 4 Skills for Self-Check, Fixes Issues, Then Comes Back to You

marsbit07/27 10:23

Agent Race Ends, Super Workbench Takes Over

The era of fragmented AI agents is ending. Over the past month, China's tech giants—Tencent, Alibaba, and ByteDance—have simultaneously shifted strategy: instead of launching new, standalone AI agents, they are consolidating their various agent projects into unified "super workbenches." Tencent integrated its QClaw teams into WorkBuddy, a strategic product hailed as a potential third flagship after QQ and WeChat. Alibaba is merging its QoderWork, Wukong, and MuleRun agents into a new "Qianwen Office" platform under DingTalk's leadership. ByteDance rebranded its TRAE SOLO coding agent to TRAE Work, signaling a broader focus on workflow collaboration. This convergence marks a pivotal industry consensus. The initial exploration phase, where companies rapidly built numerous overlapping agents for different scenarios, proved costly and inefficient. With open-source tools eroding technical barriers, competition has shifted from agent creation to resource consolidation and cost control. Historically, platform wars are won not by creating more products, but by simplifying them—as seen with browsers unifying web access and super-apps consolidating services. Now, the "super workbench" aims to become the unified AI entry point for work. This reflects a deeper market realization: the primary audience for AI is no longer just programmers (a market in the tens of millions) but all knowledge workers (a market of billions). The real opportunity lies in augmenting everyday tasks—managing emails, documents, data, and meetings—across the entire workday. The core battleground is becoming control over the primary AI entry point that employees use daily. Tencent's WorkBuddy leverages WeChat and Tencent Docs; Alibaba's Qianwen Office taps into DingTalk's organizational data; ByteDance's TRAE Work integrates with Feishu's workflows. Whoever owns this "super workbench" gains strategic control over orchestrating enterprise data and APIs. This shift is redefining enterprise software. Traditional SaaS applications, valued for their user interfaces, will recede into the background. Their core functionalities will be exposed as standardized "Skills" or APIs for the super workbench's agents to invoke. Software value will shift from selling user seats to charging based on API calls and outcomes delivered. The evolution of agents is moving through clear stages: first as novel standalone products, then as consolidated primary work entry points, and finally as pervasive, invisible capabilities embedded into the digital fabric. The recent moves by major tech firms signal the transition from the first stage into the second, accelerating toward the third. In the end, the most successful agent technology may become invisible—like electricity or the HTTP protocol—a fundamental, unnamed infrastructure powering work itself.

marsbit07/22 00:22

Agent Race Ends, Super Workbench Takes Over

marsbit07/22 00:22

Kimi Forced Two Giants to Change Pricing? Altman Rarely Admits Mistakes, Claude Resets Quotas

The article discusses the recent competitive dynamics in the AI industry, triggered by the release of Kimi's K3 model and its disruptive pricing. OpenAI's Sam Altman publicly acknowledged past shortcomings while promising an exceptional upcoming year, sparking speculation about GPT-6. Meanwhile, both OpenAI and Anthropic are engaged in a user acquisition battle, aggressively resetting usage limits and offering more credits for their coding agents (Codex/ChatGPT Work and Claude Code) to capture valuable long-task user data. OpenAI's CFO introduced a new efficiency metric: "Useful Intelligence per Dollar," arguing that true cost should be measured by completed work, not just token price. The core competition is shifting from raw model performance to which AI can seamlessly integrate into workflows as a reliable "colleague" that accomplishes tasks, with both companies launching agentic products (ChatGPT Work, Claude Cowork) designed for complex, multi-hour projects.

marsbit07/20 01:02

Kimi Forced Two Giants to Change Pricing? Altman Rarely Admits Mistakes, Claude Resets Quotas

marsbit07/20 01:02

Claude Code's Father's Night Shift AI Army, Fable 5 Built with Two Commands

For the past year, Boris Cherny, creator of Claude Code, hasn't written a single line of code himself. He manages hundreds, even thousands, of AI agents that run day and night, autonomously handling tasks from writing and submitting code to fixing CI tests and gathering user feedback. His secret lies in a "loop" system, where tasks are defined and set to run repeatedly, combined with a critical "/goal" command that provides a clear, verifiable objective and a "supervisor" model to check progress. This approach transforms the programmer's role from a coder to a system designer and reviewer. At Anthropic, AI agents even collaborate in Slack channels to divide work. The recently released Fable 5 model is built for such long-term autonomous operation, featuring self-verification and the ability to understand complex charts. Effective use requires the "/goal" command for targeted tasks and "/loop" for recurring ones. To maximize efficiency, a local context system can be set up in 20 minutes, providing the AI with persistent memory of your projects, preferences, and procedures. The key takeaway is that the future lies not in prompting AI for each step, but in building autonomous systems driven by clear, verifiable goals.

marsbit07/20 00:23

Claude Code's Father's Night Shift AI Army, Fable 5 Built with Two Commands

marsbit07/20 00:23

This Might Be the Most Stunning Image at This Year's WAIC!

This article introduces Shangtang's newly released multimodal AI model, **SenseNova U1 Pro**, unveiled at WAIC 2026. The model is highlighted for its ability to generate **native 8K-resolution images** with exceptional detail and coherence, even in extremely wide-format compositions. It goes beyond simple image generation by employing a **"图文交错思维" (interleaved image-text reasoning)** workflow, where it can plan, sketch, refine, check, and correct its outputs to achieve a final, deliverable result. Key capabilities demonstrated include generating a long 8K scroll depicting the 9-year history of WAIC, a detailed 24 solar terms illustration, a complex academic poster, a琉璃 (colored glaze)-style landscape, and a ready-to-use movie poster. The model handles intricate prompts involving layout, text clarity, material textures, and stylistic consistency. The article draws a parallel between this evolution in image generation and the progression seen in AI coding—from simple code completion to autonomous project delivery. U1 Pro represents a shift from generating single images to providing **complete content delivery systems** for professional scenarios like infographics, urban planning, and commercial design. While challenges like generation time and professional workflow integration remain, the model signifies a move towards multimodal AI agents that are accountable for producing usable, high-quality outputs.

marsbit07/19 02:39

This Might Be the Most Stunning Image at This Year's WAIC!

marsbit07/19 02:39

Claude's "Brain" Designer Just Left, Another Victory for Musk

Claude's core design lead, Jenny Wen, has announced her departure from Anthropic to join Cursor as Design Lead. This move follows the recent acquisition of Cursor's parent company, Anysphere, by SpaceXAI, effectively placing her under Elon Musk's umbrella. At Anthropic, Wen was instrumental in redesigning Claude.ai's visual interface in March 2025 and played a key role in defining the vision for the Cowork agent, shifting Claude from a chat interface to a workbench. Notably, she recently declared that traditional design processes are "dead," arguing that AI enables engineers to prototype rapidly, forcing designers to spend more time pairing with engineers and coding directly. Wen's career shift is significant. She previously left a senior design director role at Figma for an individual contributor position at Anthropic, believing hands-on experience with AI tools is crucial for leadership. For Cursor, her hiring signals a strategic pivot. As competition in AI coding tools intensifies and model advantages narrow, Cursor aims to build a moat through superior user experience, workflow, and design details—areas where Wen's expertise in crafting interfaces for AI products is seen as vital. The timing aligns with Cursor's need to evolve beyond pure engineering speed to create an indispensable tool for developers.

marsbit07/14 13:57

Claude's "Brain" Designer Just Left, Another Victory for Musk

marsbit07/14 13:57

Tsinghua AI Mathematician Emerges: From Intuition to Theorem, Contributing to an 84-Page Quantum Algorithm Paper

Tsinghua University’s Intelligent Industry Research Institute (AIR) has developed an AI mathematician agent named AIM, designed not just to solve math problems but to actively participate in early-stage research. In a recent study, researchers collaborated with AIM to develop "Sign Embedding Quantum Algorithms," resulting in an 84-page paper on quantum algorithms for matrix equations and functions. The research began with a human researcher's intuition: can rational approximation serve as a design principle for quantum algorithms? AIM helped expand this idea into multiple candidate research directions. Human researchers then filtered and focused on the most promising path. AIM assisted in organizing theorems, generating proof drafts, and performing complexity analysis, while humans maintained oversight, auditing assumptions and refining derivations. This case illustrates a human-AI collaborative workflow: AI rapidly explores and expands research avenues, generates draft materials, and aids in checking derivations; human researchers provide critical judgment on direction, value, and validity. The process emphasizes "high-throughput candidate generation + human value gating + AI-assisted audit and repair + human final integration." The resulting quantum algorithm framework offers a unified approach to several matrix problems, advancing quantum linear algebra under more general conditions. This work suggests AI's role in theoretical research is evolving from task-specific assistance to supporting the entire research lifecycle—enhancing exploration and efficiency while keeping human expertise central to guiding inquiry and ensuring rigor. *Paper & System Links:* - AIM application report: https://arxiv.org/abs/2606.24899 - Quantum algorithm paper: https://arxiv.org/abs/2604.25333 - AIM repository: https://github.com/TheoryFoundry/AIMv2

marsbit07/10 02:53

Tsinghua AI Mathematician Emerges: From Intuition to Theorem, Contributing to an 84-Page Quantum Algorithm Paper

marsbit07/10 02:53

活动图片

Technology Trendsright-arrow

Market Analysisright-arrow