2022
2022-06-21
GitHubGitHub Copilot — General Availability
Copilot becomes available to all developers at $10/month or $100/year.
2023
2023-03-14
AnthropicClaude introduced
Anthropic launches Claude and Claude Instant — a frontier assistant with coding, analysis, and writing abilities.
2023-05-17
GitHubCopilot gets better at understanding your code
Fill-in-the-middle context and richer codebase prompts roll out to Copilot.
2023-07-11
AnthropicClaude 2
100K-token context window; 71% on the HumanEval coding benchmark.
2023-08-24
Code Llama
Meta releases Code Llama — open-weight LLM specialized for code, built on Llama 2.
2023-12-14
SourcegraphSourcegraph Cody — General Availability
Sourcegraph ships Cody GA — AI code assistant across VS Code and JetBrains with Sourcegraph code-context.
2024
2024-02-27
GitHubGitHub Copilot Enterprise — General Availability
Enterprise tier at $39/user/month — codebase-aware chat, org-level knowledge base.
2024-03-04
AnthropicClaude 3 family (Opus / Sonnet / Haiku)
Three-tier Claude 3 release with a 200K context window and improved coding benchmarks.
2024-04-18
Llama 3
Meta releases Llama 3 — open-weight LLM with improved reasoning and code generation.
2024-04-29
GitHubGitHub Copilot Workspace
Copilot-native development environment (technical preview) — spec to plan to implement from natural language.
2024-05-14
GoogleGemini update at Google I/O 2024
Gemini 1.5 Flash launch, plus Gemini AI assistant features across Google developer tooling.
2024-05-29
MistralCodestral
Mistral ships Codestral — a code-specific model covering 80+ programming languages.
2024-06-21
AnthropicClaude 3.5 Sonnet
Mid-tier model that outperforms Opus 3 on coding benchmarks at 2x the speed.
2024-10-22
AnthropicClaude 3.5 Sonnet (new) + Computer Use
Upgraded 3.5 Sonnet and 3.5 Haiku; first public computer-control API.
2024-11-25
AnthropicModel Context Protocol (MCP)
Open protocol for connecting AI assistants to tools, data, and systems — spec + SDKs released.
2025
2025-01-16
CursorCursor Series B + automating-code vision
Cursor announces Series B funding round and roadmap around autonomous coding agents.
2025-02-24
AnthropicClaude 3.7 Sonnet + Claude Code preview
First hybrid reasoning model; Claude Code CLI launches in research preview.
2025-02-25
GoogleGemini Code Assist — free tier
Free tier of Gemini Code Assist for individual developers, with generous usage limits.
2025-05-19
GitHubGitHub Copilot coding agent
Asynchronous coding agent that is assigned issues and opens PRs with fixes and tests.
2025-05-20
GoogleJules (asynchronous Gemini coding agent)
Google launches Jules — an async coding agent that works in the background on delegated tasks.
2025-05-22
AnthropicClaude 4 family (Opus 4, Sonnet 4)
Flagship Claude 4 release with extended agentic-coding autonomy and stronger reasoning.
2025-05-22
GitHubCopilot agent mode — official walkthrough
GitHub publishes an Agent Mode 101 covering how Copilot agent-mode works in VS Code.
2025-08-05
AnthropicClaude Opus 4.1
74.5% on SWE-bench Verified; coding and reasoning improvements over Opus 4.
2025-09-29
AnthropicClaude Sonnet 4.5
Updated mid-tier model focused on agentic coding and long-horizon tasks.
2025-10-15
AnthropicClaude Haiku 4.5
Smaller, faster, cheaper tier of the Claude 4.x family.
2025-10-29
CursorCursor 2.0 + Composer
Unified agent-window workspace plus the Composer frontier model for multi-file code generation.
source: https://cursor.com/blog/2-0
2025-11-24
AnthropicClaude Opus 4.5
Flagship Opus 4.5 release with improved extended thinking and tool-use performance.
2026
2026-04-16
AnthropicClaude Opus 4.7 (1M context)
Opus 4.7 ships with a one-million-token context window and stronger multi-step reasoning.
Why this exists
Keeping up with AI-coding tools is like drinking from a firehose. Vendor blog posts arrive on Twitter, get buried under newer posts, and the historical arc is hard to reconstruct. This page is a compact, primary-source-cited reference: every entry points back to the vendor's original announcement so you can verify it yourself. No third-party "AI news" summaries, no marketing fluff — just dates, titles, and URLs.
If an event you expected isn't listed, it's because I couldn't find a vendor-primary URL that survives verification today. URL rot is real — this page drops entries whose primary source has 404'd rather than citing a broken link. The set of 28 events reflects what I could verify on 2026-04-22, not exhaustive coverage of every minor release.
Tools to go deeper
Kit 001
Claude Code starter kit
CLAUDE.md + 10 power prompts + hooks + agent templates.
Kit 002
Cursor Rules starter kit
Cursor rules patterns from real repos, tuned for Composer.
Kit 003
Copilot Customization starter kit
GitHub Copilot instructions files, prompt files, agent tips.
Compare
Claude Code vs Cursor
Side-by-side review of workflow and scope models.
Compare
Copilot vs Claude Code
Where cloud-first and terminal-first diverge.
Play
Copilot Config Bingo
Self-audit 25 AI-coding habits drawn from all three kits.
FAQ
Where did these events come from?
Every entry cites a primary-source URL — the vendor's own blog, news page, or changelog. I verified each URL responds with HTTP 200 on 2026-04-22 and that the page's content supports the summary shown here.
Why only 28 events? AI-coding has had way more releases than that.
Correct. This is a curated subset limited to (a) events I could verify via the vendor's primary URL on the build date, and (b) releases meaningful enough to shape how people write code. Many minor patches, pricing changes, and regional rollouts are intentionally omitted. OpenAI announcements are under-represented because openai.com rejected automated verification; rather than cite URLs I could not confirm, I left them out.
How does the date filter work?
Month precision. From 2024-01 includes every event from January 2024 onward; to 2024-12 includes every event up to and including December 2024. Leave either end blank (or click All time) to remove that bound.
What does the vendor bitmap in the URL mean?
Each vendor gets a bit in a seven-bit mask encoded as 2 hex characters. ?vendor=7f = all seven vendors on; ?vendor=01 = Anthropic only; ?vendor=02 = GitHub only. Full mapping: Anthropic=1, GitHub=2, Cursor=4, Google=8, Meta=16, Mistral=32, Sourcegraph=64. Combine with bitwise OR.
What share-URL shape does the page use?
Three query parameters: ?from=YYYY-MM&to=YYYY-MM&vendor=HH. Default (no params) = all events, all vendors. The URL bar updates live as you filter — copy at any point to share a specific view.
Does the page store anything client-side?
Only your filter state, encoded into the URL. No localStorage, no cookies, no analytics. The filter persists because the URL is the state — bookmark or share the URL to preserve it.
How do I suggest a missing event?
This is a static, human-curated page. If you find a primary-source URL (vendor blog, release notes, official docs) for a missing event, email the URL to hansng1988+timeline@gmail.com with a one-line description. If I can verify the URL and the page supports the claim, it goes in at the next build. No account, no form — just email.
Why is Claude Code not listed as its own entry?
Claude Code was announced alongside Claude 3.7 Sonnet on 2025-02-24; the primary source is the same Anthropic news post. Rather than duplicate that URL across two entries, the page treats the announcement as one event. The same reasoning applies to Claude 3.5 Sonnet (new) + Computer Use on 2024-10-22, and Cursor 2.0 + Composer on 2025-10-29.
Why is there no OpenAI row?
openai.com blocks non-browser clients with HTTP 403, which prevents this build's URL-verification step. Rather than cite URLs I could not programmatically confirm, I left OpenAI out. This is an honesty tradeoff — OpenAI has shipped major AI-coding milestones (ChatGPT Code Interpreter, Codex CLI, o1, o3, GPT-5-for-developers) that belong in any complete timeline. Future versions may include them if I can verify them via alternative means.