AS You Were
You play the route selector inside one Cisco IOS router and pick which of several candidate paths to 203.0.113.0/24 it installs. A step ladder then shows which of the first five best-path steps tied, narrowed or decided, and every wrong pick names the misconception behind it.

The builder
- Name
- Amod Darshane
- From
- Autocon6
- First spec project?
- Yes
- Began at
- NAF AC6
How it was built
- Model
- claude-opus-5-5
- Harness
- Claude Code (CLI, connected to VS Code) 2.1.326
- Spec method
- GitHub Spec Kit 1.4.3.dev0
- Spec artifacts
- constitution, spec, clarify, plan, tasks, checklists, research, data-model, contracts, quickstart
- Tasks
- 37 in 6 loops
- Built
- November 17, 2026 · 59m
- Human prompts
- 15
Build environment
- MCP servers
- claude.ai Claude Docs (connected, not called)
- claude.ai Microsoft 365 (connected, not called)
- easel (connected, not called)
- Claude Code IDE bridge for VS Code (connected, not called)
- Sources & RAG
- Workshop StarterKit CONSTITUTION.md, SPEC.md and PLAN.md, personalized by the creator
- Cisco Select BGP Best Path Algorithm, Document ID 13753
- RFC 4271 sections 4.3 and 9.1.2.2
- RFC 5398
- RFC 5737
- MDN JavaScript modules guide
- No web pages were fetched during the build; references came from the creator's inputs and model knowledge
- Skills
- Spec Kit skills: speckit-constitution, speckit-specify, speckit-clarify, speckit-plan, speckit-tasks, speckit-analyze, speckit-implement
- Other tools
- Python 3.9.6 http.server
- Google Chrome 157.0.8103.79 headless (--dump-dom, plus Python scripts over the DevTools Protocol for keyboard, layout and reduced-motion checks)
- Git
- Bash
- AI assets
- code, text and styling: claude-opus-5-5
- images: none
- audio: none
Runtime: how to play
- Type
- static-web · HTML, CSS, JavaScript ES modules, JSON
- Start
- python3 -m http.server 8000 --bind 127.0.0.1
- Then open
- http://127.0.0.1:8000
- Install needed
- No
- Internet needed
- No
- API key needed
- No
- AI at runtime
- No
- Tested on
- Google Chrome 157.0.8103.79 (headless)
- Saves
- none; progress lives in the page and reloading starts a fresh session
- Accessibility
- Keyboard-only playthrough with visible focus, Restart dialog, 360 px layout and reduced motion all checked by scripted headless Chrome. Not yet played by keyboard by a person; no screen-reader test.
Verification
- Tests
- 123 passing · python3 -m http.server 8000 --bind 127.0.0.1, then open http://127.0.0.1:8000/tests/
- Facts from
- Cisco Select BGP Best Path Algorithm, Document ID 13753, How the Best Path Algorithm Works, RFC 4271 sections 4.3 and 9.1.2.2, RFC 5398, RFC 5737
- Known limits
- Models only the first five Cisco IOS best-path steps (weight, local preference, AS-path length, origin, MED) on one router.
- Excludes eBGP over iBGP, IGP metric, oldest path, router ID and later tie-breaks, locally originated and AIGP steps, multipath, confederations, AS_SET counting, route policy, sessions and other vendors.
- bgp always-compare-med, deterministic-med and med missing-as-worst are off; every path is eBGP-learned with an explicit MED.
- Teaching text has not yet been reviewed by a person against Cisco 13753 (task T036).
- First-time-player timing and teammate explanation playtest not yet run (task T037).
- Level-summary placement follows spec FR-014; the clarify question about it was never answered.
- License
- Apache-2.0
Who it is for.
AS You Were is a browser game for network engineering students and early-career network engineers who know what BGP is but have not yet memorized the order the best-path attributes are compared in. Each round shows a table of candidate paths to one prefix, and you pick the one a Cisco IOS router installs. A correct pick shows a step ladder that marks each of the first five steps as tie, narrows, decides or not reached. A wrong pick names the misconception, such as "a shorter AS-path beats a higher local preference", and shows the values at the step you skipped. Six rounds across three levels take about ten minutes.
THE BUILD STORYHow it was made.
The creator first replaced the workshop's StarterKit SPEC.md, CONSTITUTION.md and PLAN.md with game-specific versions, then ran the Spec Kit sequence: constitution, specify, clarify, plan, tasks, analyze, implement. The creator changed direction twice early on. They dropped "No modeled winner" as a player answer, which meant both Level 3 rounds had to be redesigned so a lowest-MED decoy loses at an earlier step. They also swapped the MED values in round 3.2 so path A wins. To keep every answer identical to what a real Cisco router would install whatever order the paths arrived in, the content check also requires each winner to beat every other path one-on-one. In clarify, the creator chose to have broken content stop the game before play starts. The second clarify question, about level summaries, was left unanswered. The analyze pass found three critical issues and fixed them before any code was written. AS numbers 64520 and 64530 were outside the documentation range, which would have blocked round 2.1 from loading. Two briefings had no worked examples, and the fixtures did not cite Cisco 13753. Implementation finished 35 of 37 tasks. The test page passes 123 of 123. Deliberately breaking the engine twice caused 1 and 7 failures, and the suite passed again once each change was undone.
ADVICEFor the next builder.
- Check that the answer does not depend on arrival order: with deterministic-med off, a real router compares paths two at a time, so each winner should beat every other candidate one-on-one.
- Run analyze before implement and fix what it flags. It caught out-of-range AS numbers that would have stopped the first round from loading.
- Plan for the checks only people can do, such as reviewing the text against sources and timing a real playtest. Scripted browser checks do not replace them.
Built a game of your own?
Fill in the ARCADE template with your build agent and send me your GitHub URL.
Add your game →