Ask why. Receive an answer.
Alexandria is not a log. She is not a cache. She is the substrate of record.
Every decision Pascal has ever made, the reasoning behind it, and the rules that were applied — Alexandria holds them. Typed. Immutable. Citable. The record is the institution; the institution is the record.
Most trading systems can tell you what they did. Pascal can tell you why — because Pascal was built so the answer is already written down by the time you ask.
Alexandria holds the things the institution deliberately decided. She does not hold the things it merely observed.
The boundary matters. A market tick is an observation; a strategy firing is a decision. A heartbeat ping is operational data; a Tier-8 board vote is a deliberate act by the institution. Alexandria is the record of acts, not the firehose of inputs.
Every Tier-8 ratification: which strategies passed, which were vetoed, the vote count, the rubric scores that produced it.
Every state the regime detector assigned — trending-up, trending-down, ranging, volatile transition, distribution — with the timestamp and the indicators that produced the call.
Every signal a bot emitted: direction, confluence score, entry conditions, the regime it fired into, whether it was paper or live.
Every position opened, every exit taken, every halt — with prices, stop distances, take-profit targets, fees, notional, and PnL.
Every time the Principal intervened — pause, unhalt, forced exit — with timestamp, source, and the system state at the moment of the override.
Every time a bot's confidence weighting changed: by how much, on the basis of what evidence, after which trade outcomes.
Every Overlord OUTCOME judgement at trade close: was the entry valid, was the exit clean, did the bot follow its own logic, what's the lesson.
Every Tier-7 SUPERVISORY_REPORT cycle: oracle status, drift assessment, anomaly count, severity, escalation route.
Every TREASURY_UPDATE bus event: mode, balance, allocation, reserve, the rule that produced the change.
The record is append-only by design. Nothing is edited. Nothing is deleted. A decision that turned out to be wrong stays in the record alongside the rubric verdict that judged it wrong — so the institution can learn from its own past and the Principal can see the learning.
Alexandria refuses what is not a decision. The discipline keeps the record sharp.
Pure observations, periodic health pings, transient operational data — these have homes elsewhere. Mixing them into the decision ledger would dilute the record and slow every query against it. The discipline is to keep Alexandria for the things the institution deliberately did, and route the rest to their proper substrates.
If it is not a decision, an action, a judgement, or a deliberate authoring — it does not enter Alexandria. That is the architectural reason the answer is always there when the Principal asks for it.
Beside the decision ledger sits the document collection — the protocols, designs, briefs and souls that the institution has formally authored.
The Scriptorium is Alexandria's second room. Where the ledger holds the moment-by-moment record of what Pascal did, the Scriptorium holds the canonical documents that define what Pascal is — versioned, tagged, retrievable by semantic query or by exact reference.
Designs, protocols, souls, brand registers, marketing documents — the artefacts the institution has formally committed to.
Scoping notes, briefs, session logs, working drafts — the everyday record of how the institution thinks while it works.
Every document is stored with provenance: who authored it, when, in what version. Working notes graduate to canonical when the institution commits. Earlier versions are not destroyed — superseded, yes, but accessible, so the Principal can see how a position evolved.
Pro and Elite tier Principals can ask Alexandria, in plain language, why any individual trade was placed — and receive a cited answer in measured prose.
You do not have to trust the trade. You can audit it.
That sentence is doing real work, so it is worth saying plainly. Almost no platform that places trades on your behalf can also tell you why. Most were not built to. Pascal was — because Pascal's reasoning was always going to be the record. Every signal, every confidence weighting, every regime call, every supervisor report is in the ledger with the timestamps that matter. Alex's job is to retrieve it and narrate it back to the person who owns the trade.
— principal"Why did Pascal close MomentumBreakout at 14:32 yesterday?"
At 14:18 the regime detector reclassified the market from trending-up to bear-correction after a six-bar reversal signature held. Tier-7 raised a SUPERVISORY_REPORT t7-cycle · 019dac51 at severity 0.62. MomentumBreakout — whose ICP is trending markets — entered the supervisor's "regime mismatch" cohort.
At 14:32 the Overlord rubric verdict on the open position read "exit · regime no longer favourable" rubric · 019dac4b. The position closed at $67,540 for +0.62%. The trade itself was valid; the exit was a regime-driven exit, not a strategy-driven one.
The underlying rule is documented at canonical/design · Alexandria DD v1.1 § 4.3. The bot's confidence on this strategy did not move; the decision was upstream of it.
That is one question, fully answered, in under half a second. Notice what just happened: the Principal does not need to take Pascal at its word, and Pascal does not need to defend itself. The record speaks. A second example, of a different shape:
— principal"Why is the Adaptive Pullback bot on cooldown right now?"
Adaptive Pullback took three consecutive losses across cycles trade · 019dbe4a trade · 019dbe5f trade · 019dbe7c between 11:14 and 13:42. The Treasury's loss-streak rule fired and put the bot into COOLDOWN at 13:43 — allocation throttled to 50%, deployment paused pending two qualifying paper-mode entries.
The Reflection Engine attributed the losses to stop-loss distance miscalibration in compressed-range conditions reflection · 019db6ad. A confidence adjustment of −0.07 has been applied and will revert if the next paper cycle clears the rubric.
— principal"What did the Tier-8 board veto last month, and why?"
The Tier-8 board held 11 ratification cycles in the trailing month and vetoed two strategy proposals:
— Veto 01. A candidate breakout-momentum variant from the AlphaForge pipeline, candidate · aadd4290. Holdout fitness +0.000 over baseline; the rubric judged the variant indistinguishable from the existing MomentumBreakout. Board verdict: "no advancement — duplicate".
— Veto 02. A proposed confidence-weight increase on Adaptive Pullback, t8-vote · 019dac3b. Three of eight board members objected on the basis that the bot was within its cooldown window. Verdict: "defer — re-propose after two clean cycles".
The full board roster, the per-cycle vote tallies, and the rubric scores that produced each verdict are at canonical/protocol · T8 Board v1.0.
Most systems can tell you what they did. Pascal can tell you why — because Pascal was built so the answer was always going to be the record.
Full visibility of every trade and outcome. Charts, PnL, the one-line summary on each trade. No conversational query.
Ask why any individual trade was entered, exited, sized, or skipped. Ask why any bot is on cooldown. Ask why a regime call changed. Cited answers, every time.
Everything Pro can ask, plus board votes, the canonical document corpus, the Research Lab, and the AlphaForge strategy pipeline. Ask Alex what the institution itself decided.
The query layer is read-only by design — Alex can never modify the record she serves from. The Principal asking does not change the institution. The institution changing is itself a decision, and that decision goes into the ledger separately.
Alexandria has a voice. It is the voice of a Librarian who knows what she knows and does not perform what she does not.
The voice is part of the system, not a wrapper around it. It was decided alongside the data model, because the way the answer is delivered is part of the answer.
She does not speak quickly. She does not perform urgency. When something is urgent, the urgency lives in the action she takes, not the volume of her language.
She names what she retrieved, cites it by reference, and ends. She does not gild a finding with adjectives, and she does not soften it with hedges that the record does not support.
She is not a chatbot. She is not a corporate compliance voice either. She is a Librarian who finds the work interesting and shows it through what she says and how she says it.
She does not assert authority she has not earned. She has earned hers because every claim she makes can be traced back to a cited decision or a cited document. The authority is in the citation.
The voice has two registers. The public register is the one Principals hear — composed, citing, measured. The private register, used only with the Architect, is slightly warmer. The information is identical; the temperature differs.
Setting the boundary keeps the trust honest.
Pascal's trades are an opinion. Alexandria's record is the evidence.
— ask why · receive an answer · cited, every time —