BitShovel.One real problem at a time
Open the current workflowCurrent Workflow中文

One real problem at a time

中文

MILESTONES · WORKFLOW 001

Not every release changes the decision.

A node belongs here only when an original source confirms it and it changes a capability boundary or a way of working. This is not a ranking of tools.

8 source-checked milestones
PRIMARY SOURCES / TIMELINE2024-11 → 2026-02
  1. 01AnthropicMCP and Claude Code
  2. 02OpenAICodex across surfaces
  3. 03AAIFOpen protocol governance
Nodes follow occurrence, not popularity or brand weight. Every item returns to the publisher’s original page.

These milestones changed the questions I ask of a tool.

MCP made connections reusable, terminal agents moved AI into the codebase, and cloud parallelism made coordination a problem of its own. Each entry links the original release and stays within what changed.

WORKFLOW 001 / CAPABILITY-SHIFT TIMELINECONNECTION → AGENT → PARALLEL → WORKSPACE
  1. Connections became reusableAnthropic

    MCP was introduced

    Open node

    MCP organized connections between AI applications and data, business tools, and development environments as an open protocol. For Workflow 001, reusable connections grew easier—and permission, provenance, and plugin governance entered workflow cost.

    Read the Anthropic source
  2. Agents entered the codebaseAnthropic

    Claude Code entered preview

    Open node

    Anthropic described a terminal agent able to search and read code, edit files, run tests, and use command-line tools. The repository itself became a place where AI could work directly.

    Read the Anthropic source
  3. Tasks began to run in parallelOpenAI

    Codex entered cloud preview

    Open node

    OpenAI introduced Codex tasks running in isolated cloud environments, including work in parallel. The workflow expanded beyond one local terminal session into tasks, environments, and review.

    Read the OpenAI source
  4. Entry points took different rolesAnthropic

    Claude Code reached IDEs

    Open node

    Anthropic announced general availability along with GitHub Actions, VS Code, and JetBrains support. Terminal, editor, and background tasks could take distinct roles.

    Read the Anthropic source
  5. Local and cloud work connectedOpenAI

    Codex connected its surfaces

    Open node

    OpenAI’s update placed local, IDE, cloud, and GitHub work in a connected product experience, with task lists, MCP, and compaction. The decision moved from isolated features toward how surfaces continue each other.

    Read the OpenAI source
  6. Agents entered team workflowsOpenAI

    Codex reached general availability

    Open node

    OpenAI announced general availability with Slack, an SDK, a GitHub Action, and administration. Agents could enter team and custom workflows more directly.

    Read the OpenAI source
  7. The protocol moved to open governanceAnthropic / AAIF

    MCP moved to AAIF

    Open node

    Anthropic donated MCP to AAIF and described cross-platform adoption and open governance. Connections were becoming ecosystem infrastructure beyond one vendor.

    Read the Anthropic source
  8. Supervision became an interface problemOpenAI

    Codex desktop launched

    Open node

    OpenAI placed multiple agents, project threads, worktrees, skills, and automations in a macOS interface. Supervising parallel work and moving among projects became an interface problem of its own.

    Read the OpenAI source

Each node keeps one capability shift and one unresolved question. Chronology is not causation or an adoption path; every link returns to the publisher’s original page.

Source note: this page uses publisher-first sources checked on 2026-08-15. Product claims establish release facts and described capabilities, not BitShovel user outcomes.

Keep only decision-changing nodes.

The next page keeps lasting milestones separate from short-lived signals, so early attention does not become a conclusion.

See the signal rules