DeepSeek DSH Superagent: From GitHub Requests to Agent Workflows
Understand how DeepSeek's DSH Superagent prototype turns authorized GitHub and Feishu requests into coding tasks, and which workflow mechanisms you can adapt.
8 min
AI agents, models, and workflow guides
Browse the complete Floatboat library: model analysis, agent architecture, automation workflows, file tools, and practical systems for solo operators.
171 articles
OpenAI GPT-6 Cyber: Who Gets Access and What It Can Do
OpenAI’s GPT-6 Cyber starts with a limited Daybreak Red rollout. Check eligibility, Codex and API access, and token pricing, then see how to interpret the benchmarks and scope a defensive review.
Kostja11 min read
Browse the library
170 articles
What Is Gemini Agent? Google's Universal Work Agent Explained
Gemini Agent is Google's universal work agent for persistent tasks, multi-agent coordination, model routing, and work across business tools. Here's what is confirmed, what remains in preview, and how it differs from the Gemini app.
12 min
AI Agent Connectors Explained: MCP, OAuth, and Permissions
Learn how AI agent connectors use MCP, OAuth, scopes, resource permissions, and approval gates—and how to evaluate a connector before it can read or change your work.
13 min
Local-First vs Cloud Agent Workspaces: Which Fits?
Compare local-first and cloud agent workspaces across files, privacy, background execution, connectors, reliability, and control—then choose the right architecture.
9 min
What Is an Agent Harness? Architecture, Loops, and Safety
Learn what an AI agent harness does, how its execution loop, context, tools, sandbox, approvals, tracing, and recovery fit together, and how to evaluate one.
13 min
What Is ChatGPT Space? Pages, Projects, and Permissions
ChatGPT Space organizes editable Pages, files, and collaboration. Learn how it differs from Projects and chat, how permissions work, and when to migrate.
12 min
GPT-6.1 Sol: Pricing, Benchmarks, Availability, and How It Compares
GPT-6.1 Sol brings near-Astra coding, computer use, and professional-work performance at one-fifth of Astra's token prices. Here is what the numbers do—and do not—mean.
10 min
OpenAI DevDay 2026: Every Major Announcement, Explained
A complete map of OpenAI DevDay 2026: Dots, GPT-6.1 Sol, Codex Cloud, Agents API, Plugins, Space, Pages, Pro 500, and the platform strategy connecting them.
11 min
OpenAI Plugins Explained: Skills, MCP, UI, Events, and Distribution
What OpenAI Plugins are in 2026, how skills, MCP servers, UI extensions, events, and hooks fit together, and when a smaller integration is the better choice.
10 min
Batch Convert Markdown Files — When One File Becomes a Folder
How to batch convert a folder of markdown files to PDF, Word or HTML: pandoc scripts, GUI tools, failure modes, and when folder-scale work needs an agent.
14 min
Best Markdown Editors in 2026 — Ranked by Workflow, Not Features
Obsidian, Typora, VS Code, Zettlr, iA Writer and more, ranked by workflow fit. What each editor trades away, and where browser tools beat installed apps.
16 min
Math in Markdown: LaTeX Syntax and Where It Renders
How math in markdown works: dollar-sign LaTeX syntax, where it renders (GitHub, Obsidian, Jekyll), the KaTeX vs MathJax split, and honest workarounds.
13 min
From these guides to your next workflow
Route events to the right model tier
Complex meeting prep goes to the frontier tier, routine classification to the budget tier — one consistent behavioral family.
Prep, execute, follow up — from your calendar
Agents read the event, pull context, draft the brief, and send the follow-up. No prompt engineering, no app switching.
Continuous operation at SaaS prices
Budget-tier models make always-on classification viable. Subscribe to a tournament, watch the updates arrive — no budget anxiety.
Frequently Asked Questions
Where should I start if I am new to agent workflows?
Begin with the AI agents category to distinguish models, agents, harnesses, and workspaces. Then follow one workflow guide that matches the files and deliverable you actually work with.
How do I find an implementation guide rather than a product announcement?
Use the topic tabs to narrow the library, then read the description before opening an article. Look for a concrete task, inputs, execution steps, and verification; a release analysis may explain a capability without providing setup instructions.
Are model comparisons permanent recommendations?
Treat a comparison as a dated analysis. Provider access, model IDs, prices, and supported tools can change; check the article date and its linked sources before using it for a new purchasing or deployment decision.
What is the difference between this blog and the Skills Marketplace?
The blog explains concepts, experiments, and workflows. The Skills Marketplace organizes publishers, open-source projects, and Skills with their source links. An article may link to a project page when that project is part of the explanation.
Can I use these workflows outside Floatboat?
Many articles discuss general methods such as source checking, task boundaries, and deliverable review. Separate those methods from instructions that depend on a named product, connector, or execution environment.
How do I follow an article between English and Chinese?
Use the language switch on the article page to open its counterpart. Keep the underlying source links and publication context in view when comparing terminology; language choice does not change the release date being discussed.