Anthropic launches its latest model, Claude Opus 4.8! Claude Code simultaneously adds "Dynamic Workflows" — one person can match a hundred-person development team

📄Full Article· Automatically extracted by trafilaturaGemini 翻譯1887 words

The ultimate AI colleague for programmers has arrived! Anthropic announced today (28th) the release of the Claude Opus 4.8 update, which not only enhances the model's judgment and long-horizon independent working capabilities at the original price, but also introduces "Fast mode" — 2.5x faster with prices slashed to one-third. In addition, the company simultaneously unveiled an epic new feature, "Dynamic Workflows," enabling Claude to act as a project manager, directing hundreds of AI Subagents to work in parallel and self-verify automatically. The well-known project Bun put it to the test, successfully porting 750,000 lines of code from Zig to Rust in just 11 days! (Previous context: Anthropic's revenue overtakes OpenAI: A historic reversal in the AI arms race) (Background supplement: The era of the one-person company has arrived! Anthropic releases "AI-Native Startup Handbook": You can be a tech CEO even without knowing how to write Code) The application of artificial intelligence in the software engineering field has officially crossed from "assisted coding" into a new era of "fully automated project outsourcing." AI unicorn Anthropic took a two-pronged approach today (May 28, 2026), not only announcing the release of the latest Claude Opus 4.8 model on social platform X, but also simultaneously revealing on its official blog a killer preview feature poised to overturn existing software development workflows: Dynamic Workflows. Introducing Claude Opus 4.8: it builds on Opus 4.7 with sharper judgment, more honesty about its own progress, and the ability to work independently for longer than its predecessors. Available today at the same price. pic.twitter.com/EufxL7T1kb — Claude (@claudeai) May 28, 2026 According to the official Claude (@claudeai) announcement, Claude Opus 4.8 builds upon its predecessor, bringing sharper logical judgment, more honest assessment of its own progress, and longer-duration "independent working capability" than the previous generation. Notably, the new model maintains the same pricing as before. At the same time, Anthropic has prepared a brand-new Fast mode for developers. In this mode, users can run the same model at 2.5x the speed, with prices dramatically reduced to one-third of the original. Developers only need to enter /fast in the Claude Code terminal to switch with a single command. If the Opus 4.8 upgrade is about polishing the tool, then the newly launched "Dynamic Workflows" (currently in research preview) in Claude Code completely transforms the way of working. In the past, a single AI agent struggled to handle extremely large and complex engineering tasks in one go (such as framework replacements or API deprecations spanning hundreds of files). Now, when users hand over high-difficulty tasks, Claude takes over everything "End-to-end": - Dynamic planning and Fan out: Claude automatically writes coordination scripts, breaks down large tasks, and launches dozens to hundreds of parallel Subagents to work simultaneously in one shot. - Adversarial and self-verification: Each agent tackles the problem from different angles, and "Adversarial agents" are even deployed to try to break existing results, continuously iterating until convergence. - Long-running execution: Supports uninterrupted execution from hours to days, and can resume from saved progress even if interrupted. To demonstrate the strength of this technology, the company disclosed a stunning real-world case. Jarred Sumner, author of the well-known JavaScript runtime Bun, used Dynamic Workflows to fully port Bun's underlying language from Zig to Rust. This massive engineering effort, which typically requires an entire engineering team months or even years, under Claude's direction: - Claude assigned hundreds of agents to write .rs files in parallel, with each file equipped with two AI reviewers. - Continuously drove compilation and testing through a Fix loop until everything fully passed. - Final result: Processed approximately 750,000 lines of Rust code, with a test pass rate as high as 99.8%, and from the first commit to the final merge, it took only 11 days! The company reminds that since launching hundreds of subagents consumes significantly more Tokens (compute costs), developers are advised to start testing with smaller-scoped tasks initially. Users can experience this powerful multi-agent collaboration feature through direct commands or by enabling ultracode in the settings.

Data Status✓ Full text extractedRead Original (動區 BlockTempo)

🔍Historical Similar Events· Keyword + Asset Matching6 items

2026-04-17

Anthropic Debuts Claude Opus 4.7 as Agentic Workflows Take Center Stage

2026-04-21

Google assembles elite team to catch up with Anthropic, even its own engineers are secretly using Claude Code

2026-05-28

Anthropic's Claude Opus 4.8 Is Here: Better AI Coding, Smarter Safety—Same Huge Price

2026-05-28

Claude Opus 4.8 Rolls Out: Anthropic Strikes Back in AI Race

2026-05-26

Two months after Claude Code burned through Uber's annual budget, the COO bluntly stated: There is no proportional relationship between Token consumption and useful output.