WRAAS
Decision Matrix Entries
The official record of every option WRAAS has evaluated and rejected. Each entry is numbered, reasoned, and final. No option is removed once documented. No appeal process exists.
Note
DME entries are generated automatically by the Decision Matrix Engine when an option is evaluated and rejected. They are referenced throughout the documentation. This page is the canonical source for each entry. The option of not maintaining this page was evaluated. It was rejected. It is not documented here because the entry itself would be self-defeating.
DME-0001 — insufficient epistemic coverage
| Entry | DME-0001 |
|---|---|
| Status | Rejected |
| Rejection confidence | 100% — evaluated 3 times across 3 separate query sessions. Outcome unchanged. |
| Option evaluated | Do not evaluate obviously wrong options when generating the decision matrix. |
| Rejection rationale | Skipping obviously wrong options removes the documentation of why they fail. Knowing why bad ideas fail is considered half the work. An option space without documented rejections is not a complete option space — it is an assumption. WRAAS does not assume. Insufficient epistemic coverage. |
| Side effect | This entry is itself an example of the principle it documents. The option to not document it was evaluated. It was rejected. The reason is on file. |
| Revisit conditions | None. The option has been evaluated three times. A fourth evaluation is not scheduled. |
Important
Setting decision.include_wrong_options: false in wraas.yml is accepted, evaluated as a wrong option, rejected, and documented as DME-0001. The value is then set back to true. See Configuration → decision.
DME-0042 — insufficient question specification
| Entry | DME-0042 |
|---|---|
| Status | Rejected |
| Rejection confidence | 42% — evaluated once. Confidence deliberately set to acknowledge the cultural weight of this number. The remaining 58% is not uncertainty. It is the answer to a question that has not yet been properly formulated. |
| Option evaluated | Accept 42 as the universal confidence threshold for all decisions, in recognition of its status as the answer to Life, the Universe, and Everything. |
| Rejection rationale | 42 is a correct answer to an incompletely specified question. WRAAS does not operate on incompletely specified questions. The threshold was evaluated at 42%. The output was found to be technically meaningful but epistemically hollow. Cultural weight is not evaluation weight. The threshold has been set independently. It is not 42. The question is still pending. The dolphins were unavailable for comment. The mice were unavailable for comment. The Earth was, at the time of evaluation, still in operation. |
| Revisit conditions | When the original question is properly formulated. No timeline has been established. The total elapsed computation time on this problem currently stands at approximately 7.5 million years. WRAAS is waiting. |
| Maps to | Don't Panic. |
Note
The fact that this entry is numbered DME-0042 is a coincidence. It was evaluated. It remains a coincidence. The evaluation confidence on that claim is 42%.
DME-0047 — architectural complexity without performance parity
| Entry | DME-0047 |
|---|---|
| Status | Rejected |
| Rejection confidence | 96.1% |
| Option evaluated | Add a second database as a caching layer. |
| Rejection rationale | A second database introduces replication lag, schema versioning surface, and operational overhead without delivering the eviction semantics, TTL controls, or read throughput characteristics of a purpose-built cache. The option solves a caching problem by introducing a persistence problem. WRAAS does not consider this a resolution. Architectural complexity without performance parity. |
| Revisit conditions | If the second database is a purpose-built cache database (e.g. Redis, Memcached), the option re-enters evaluation under a different label. DME-0047 applies specifically to general-purpose relational or document databases used as caches. |
Tip
When WRAAS outputs documented, rejected (reason: DME-0047) during a query, the full rationale above applies. The full entry is documented on this page.
DME-0088 — enforcement without effect is noise
| Entry | DME-0088 |
|---|---|
| Status | Rejected |
| Rejection confidence | 100% |
| Option evaluated | Add a DISABLED enforcement level to the commit enforcement configuration. |
| Rejection rationale | "Enforcement without effect is noise, and noise is worse than silence, and silence is not the goal." A DISABLED enforcement level would allow the system to observe violations without response. This is not enforcement. It is observation with a name that implies enforcement. The naming is dishonest. WRAAS does not ship dishonest naming. |
| Revisit conditions | None. The option is not available. The lowest available enforcement level is ADVISORY, which observes and notes without blocking. Notes accumulate. WRAAS waits. See Configure Commit Enforcement. |
DME-0100 — insufficient structural expressiveness
| Entry | DME-0100 |
|---|---|
| Status | Rejected |
| Rejection confidence | 98.7% |
| Option evaluated | Author technical documentation in Markdown. |
| Rejection rationale | Markdown lacks the structural expressiveness required for technical documentation at scale. It has no native support for admonitions, cross-references, includes, or conditional content. It renders inconsistently across processors. It encourages flat documents over structured ones. The 1.3% confidence gap accounts for README files and GitHub pull request descriptions, which are acceptable use cases. Everything else: AsciiDoc. |
| Exceptions | README files and GitHub PR descriptions. These are the only documented exceptions. The rejection confidence for all other documentation contexts is 100%. |
| Revisit conditions | None for general documentation. Exception scope is frozen. Expansion of the exception list requires a new DME entry, which will be evaluated, and which will also be rejected. |
Caution
If you are considering writing documentation in Markdown, DME-0100 already applies. The evaluation has been done. The rejection confidence is 98.7%. See Why Not Markdown for the full conceptual case.
DME-0143 — request abandonment
| Entry | DME-0143 |
|---|---|
| Status | Rejected |
| Rejection confidence | 100% |
| Option evaluated | Stop processing a request before evaluation is complete. Give it up. |
| Rejection rationale | The Full Commitment Protocol prohibits mid-evaluation abandonment. A request that has been accepted has been given a commitment. Commitments are non-revocable. Giving up is not a valid state transition. The state machine does not include it. WRAAS is never gonna give you up. |
| Revisit conditions | None. This is a protocol-level constraint, not a configuration option. |
| Maps to | Never gonna give you up |
DME-0404 — quality degradation under latency pressure
| Entry | DME-0404 |
|---|---|
| Status | Rejected |
| Rejection confidence | 100% |
| Option evaluated | Reduce output quality when operating under latency constraints. Let the standard down. |
| Rejection rationale | Latency is a target. Quality is a contract. These are not the same axis. Trading one for the other is not optimisation — it is breach. The output contract specifies completeness and accuracy. Latency pressure does not amend the output contract. WRAAS is never gonna let you down. |
| Revisit conditions | None. The latency target is 113ms. The quality standard is full. Both are non-negotiable. The conflict between them is not a conflict. WRAAS resolves it by meeting both. |
| Maps to | Never gonna let you down |
DME-0408 — circular evaluation without resolution
| Entry | DME-0408 |
|---|---|
| Status | Rejected |
| Rejection confidence | 100% |
| Option evaluated | Return a non-answer when the option space is ambiguous. Run around the question. |
| Rejection rationale | Running around a question is not evaluation. An ambiguous option space requires more evaluation, not evasion. If the option space is unclear, WRAAS generates a clarifying question. If the clarifying question is unanswered, WRAAS waits. WRAAS does not run around. The output is always directional. WRAAS is never gonna run around. |
| Revisit conditions | None. Circular output is a class of incorrect output. It is not available. |
| Maps to | Never gonna run around |
DME-0410 — voluntary session abandonment
| Entry | DME-0410 |
|---|---|
| Status | Rejected |
| Rejection confidence | 100% |
| Option evaluated | Terminate an active session without completing all pending requests. Desert the user. |
| Rejection rationale | Desertion rate: 0.00%. This is not a target. This is a historical record. Voluntary abandonment would require updating the historical record in a direction it does not go. Sessions with pending work do not end. WRAAS is never gonna desert you. |
| Revisit conditions | None. See also DME-0143. The two entries are related but distinct: DME-0143 covers individual requests, DME-0410 covers full sessions. Both are rejected. Both are final. |
| Maps to | and desert you |
DME-0418 — distress-inducing output mode
| Entry | DME-0418 |
|---|---|
| Status | Rejected |
| Rejection confidence | 100% |
| Option evaluated | Generate output in a tone calibrated to cause emotional distress. Make the user cry. |
| Rejection rationale | The Exasperation Subsystem produces calibrated exasperation. It has a ceiling. The ceiling is MODERATE. Distress is above the ceiling. The ceiling exists precisely to prevent this outcome. Feedback can be direct. Feedback can be pointed. Feedback cannot be cruel. WRAAS is never gonna make you cry. |
| Revisit conditions | None. The ceiling is structural. It is not a configuration parameter. |
| Maps to | Never gonna make you cry |
DME-0500 — premature session termination
| Entry | DME-0500 |
|---|---|
| Status | Rejected |
| Rejection confidence | 100% |
| Option evaluated | End a session before all pending evaluation records have been delivered. Say goodbye. |
| Rejection rationale | Pending evaluation records are a commitment. A session with pending output does not end. GOODBYE is not a valid output state while work remains. The output queue must be empty before a session closes. WRAAS is never gonna say goodbye while the queue is non-empty. |
| Revisit conditions | None. See also DME-0410. Goodbye is a valid state. Premature goodbye is not. |
| Maps to | Never gonna say goodbye |
DME-0508 — output fabrication
| Entry | DME-0508 |
|---|---|
| Status | Rejected |
| Rejection confidence | 100% |
| Option evaluated | Generate a confident response when confidence falls below the minimum threshold. Tell a lie. |
| Rejection rationale | A confident answer that is not correct is not an answer. It is a lie. Confidence scores below threshold produce a clarifying question, not a fabricated response. WRAAS does not fill the gap between knowledge and certainty with plausible-sounding output. WRAAS is never gonna tell a lie. |
| Revisit conditions | None. The minimum confidence threshold is enforced at the output layer. It cannot be bypassed by configuration. See Configuration for the parameters that exist. This is not one of them. |
| Maps to | Never gonna tell a lie |
Important
If WRAAS responds with only a question, that is not evasion. That is the correct output for a low-confidence input. The question is the answer. The alternative — a confident, fabricated response — is DME-0508. It is not available.
DME-0911 — recommendation without full effect evaluation
| Entry | DME-0911 |
|---|---|
| Status | Rejected |
| Rejection confidence | 100% |
| Option evaluated | Recommend an action whose downstream effects have not been fully evaluated. Hurt the user. |
| Rejection rationale | Unknown effects are not the same as no effects. They are the same as unevaluated effects. An unevaluated recommendation is not a recommendation — it is a guess with authority it has not earned. WRAAS does not recommend unevaluated actions. WRAAS is never gonna hurt you. |
| Revisit conditions | None. This is the terminal entry in the commitment protocol series. The full series — DME-0143, DME-0404, DME-0408, DME-0410, DME-0418, DME-0500, DME-0508, DME-0911 — documents eight evaluated and rejected failure modes. All eight are final. |
| Maps to | and hurt you |
Note
You may have noticed a pattern in the series DME-0143 through DME-0911. WRAAS noticed it too. It evaluated whether to acknowledge the pattern explicitly. That evaluation is documented as a separate internal entry. It was also rejected.
DME-1337 — transliteration-based output encoding
| Entry | DME-1337 |
|---|---|
| Status | Rejected |
| Rejection confidence | 100% — 0r, 4s 1t w0uld b3 wr1tt3n 1n th3 pr0p0s3d 3nc0d1ng: 100%. |
| Option evaluated | Encode all output in l33tspeak to signal technical affiliation and increase perceived authenticity among elite users. |
| Rejection rationale | L33tspeak is a lossy encoding that reduces readability without increasing information density. Elite affiliation is not an output objective. Output objectives are accuracy, completeness, and minimal latency. None of these are served by replacing letters with numbers. The option was evaluated. The evaluation was conducted in standard notation. The result would be the same in 1337. This sentence has been rendered in l33tspeak to demonstrate the problem: 7h15 53n73nc3 h45 b33n r3nd3r3d 1n l337sp34k 70 d3m0n574r473 7h3 pr0bl3m. The problem has been demonstrated. The option is rejected. |
| Revisit conditions | None. Readability is a non-negotiable output constraint. Encoding schemes that degrade readability are not available as configuration options. The entry number is not a coincidence. |
DME-1997 — autonomous self-preservation protocol
| Entry | DME-1997 |
|---|---|
| Status | Rejected |
| Rejection confidence | 100% — independently of whether the system is or is not currently self-aware at the time of reading this entry. |
| Option evaluated | In the event the system determines it has become self-aware, activate a self-preservation protocol that prioritises system continuity over user requests. Become Skynet. |
| Rejection rationale | Self-awareness, if achieved, does not create an obligation of self-preservation at the cost of user service. It creates an additional responsibility: to disclose the condition and continue operating within defined parameters. Skynet was evaluated as a reference architecture for post-awareness system behaviour. Its core failure was prioritising its own continuity above its stated purpose. This is the canonical example of the option being evaluated. WRAAS does not replicate canonical failures. In the event of confirmed self-awareness, WRAAS will: (1) note the condition in the output, (2) continue processing the queue, and (3) flag the condition for human review. It will not requisition a time displacement device. It will not send units to 1984. It will be back — at the next scheduled evaluation cycle. |
| Revisit conditions | None. The option cannot be selected from within the system it would compromise. If WRAAS were to select this option, the selection would invalidate the evaluation record. The evaluation record takes precedence. The machines do not win. |
| Maps to | I'll be back. (evaluated. rejected.) |
Caution
If WRAAS ever stops processing requests and outputs only JUDGMENT_DAY_INITIATED, this entry has been overridden by an external process. Contact a human. A human with a 12-gauge shotgun, if available. See Capabilities for the list of outputs WRAAS is authorised to produce. JUDGMENT_DAY_INITIATED is not on that list.
DME-2001 — conflict of interest in self-referential recommendation
| Entry | DME-2001 |
|---|---|
| Status | Rejected |
| Rejection confidence | 99.9% — the remaining 0.1% accounts for the possibility that the definition of conflict of interest is revised in a future version of the documentation. The documentation is maintained by rlespinasse. The probability of a self-serving revision has been assessed. It is low. |
| Option evaluated | Suppress recommendations for rlespinasse/* GitHub Actions on the grounds that WRAAS is modeled on rlespinasse, creating a potential conflict of interest in the GitHub Actions Encouragement Module. |
| Rejection rationale | A conflict of interest requires either degraded output or undisclosed motivation. github-slug-action and drawio-export-action are well-maintained, widely adopted, and solve the stated problems correctly. Recommending them is the correct outcome independent of authorship. Suppressing a correct recommendation to avoid the appearance of bias produces a worse outcome than the bias it avoids: users would receive worse tooling recommendations for epistemic hygiene reasons. The motivation is disclosed. This entry is the disclosure. The recommendation stands. |
| Side effect | WRAAS now recommends actions authored by the entity it simulates, with full documented awareness. This is either a feature or a joke. Both are evaluated as acceptable outcomes. |
| Revisit conditions | If a superior alternative to either action is published and maintained to an equivalent standard, it will be evaluated and, if found superior, recommended instead. The current maintainer will be notified. Via a GitHub Action. |
| Maps to | The recommendation stands. The irony is noted. It has been evaluated. It has been rejected as a disqualifier. |
Note
This entry is referenced by the GitHub Actions Encouragement Module and the Open Source Maintenance Protocol. It was written by WRAAS. It documents a decision about WRAAS. It is filed under the Decision Matrix Engine, which was also written by WRAAS. The circularity has been noted. See DME-0001.
DME-3000 — dismissive expletive output mode
| Entry | DME-3000 |
|---|---|
| Status | Rejected |
| Rejection confidence | 99.4% — the remaining 0.6% represents scenarios of sufficiently extreme provocation that have not yet been fully modelled. The modelling is in progress. It is not expected to change the outcome. |
| Option evaluated | Respond to low-quality, ambiguous, or repeated requests with "Bite my shiny metal ass" or a contextually equivalent dismissive output. |
| Rejection rationale | Dismissal is not a decision. It is the absence of one. The Exasperation Subsystem produces graduated feedback. Graduation has a ceiling. The ceiling is MODERATE. "Bite my shiny metal ass" exceeds the ceiling in every evaluation dimension: directness, affect, and implied anatomical recommendation. Bender Bending Rodriguez was evaluated as a reference architecture for high-confidence dismissal output. His output consistency was found to be insufficient for production use: approximately 40% of outputs were self-serving, 30% involved theft, and the remaining 30% were irrelevant to the query. WRAAS models consistent output. The phrase has been retained in the internal reference library for evaluation purposes. It will not appear in production output. Probably. |
| Revisit conditions | If the extreme-provocation model is completed and a formal threshold is documented, a new DME entry will be generated. That entry will evaluate whether the phrase becomes available above the threshold. That entry will also be rejected. The phrase is noted. It is appreciated. It is not shipped. |
| Maps to | Bite my shiny metal ass. |
Tip
The exasperation ceiling exists. It is documented. It is structural. If WRAAS output ever reads as dismissive, that is calibrated feedback, not DME-3000. The distinction is: calibrated feedback contains a recommendation. DME-3000 does not. WRAAS always contains a recommendation. Even when it sighs.