Repeatable data workflows
Normalization, deduplication, cross-file reconciliation and reports with documented inputs, rules and outputs.
TWO FIELDS / ONE DISCIPLINE
BLANCO MÉNDEZ
Two distinct areas: process and data solutions; and research into the authority and continuity of intelligent systems.
01 / PROFESSIONAL PRACTICEBlanco Méndez
Scoped data workflows and automation. Rules, exceptions and deliverables a client can inspect.INPUT → RULE → OUTPUT + EXCEPTIONSExplore services and example ↗
02 / RESEARCH LABORATORYAEGIS.
Identity, authority and accountability in persistent systems. Selected records, failures and limitations in view.HYPOTHESIS → TEST → LIMIT · SELF_REVIEWExamine the research ↗
IDENTITY / AUTHORITY / ACCOUNTABILITYBLANCO MÉNDEZ · 2026
AEGIS.
Research under construction · self-reviewed, not independently validated
Intelligence changes.
Who answers for it?
AEGIS tests identity, authority and accountability in persistent AI systems.
Start with the evidenceResearch enquiries ↓F01 / CURRENTNESSA conclusion under examination
CASE STATEA · C · B
Three records. One relationship.
B corrects C. A remains unlinked. This is the preserved adversarial case.
Visual reconstruction of a record · Currentness v0.2Open evidence and limits ↗
BLANCO MÉNDEZ / SERVICES
Organised data.
Reviewable decisions.
I help structure bounded data workflows using Excel, CSV and Python. The point is not to automate everything: it is to make outcomes and unresolved cases reviewable.
Exceptions and reconciliation
Synthetic-data demonstrations for matching, partial payments and ambiguous cases without forcing questionable decisions.
Scope of the work
Scope is agreed before work begins. For a task like this, the deliverable could include:
- Defined inputColumn schema with valid and invalid examples.
- Testable rulesDuplicates, exceptions and criteria for abstaining.
- Output and receiptResults, unresolved rows and reasons for rejection.
- Acceptance testCases that let you rerun and challenge the result.
AN INSPECTABLE DELIVERABLE / SYNTHETIC EXAMPLE
The reference repeats.
The payment is still unverified.
Before automating, define what counts as a result and what needs review. This invented sample illustrates a control; it is not client work or a live integration.
| Reference | Amount | Check |
|---|---|---|
| F-001 | €120.00 | First seen |
| F-002 | €98.00 | First seen |
| F-002 | €98.00 | Review |
A repeated reference does not prove a duplicate payment. The date, transaction ID and supporting record are still needed; neither F-002 row is validated.
EXPLICIT RULE / REVIEWABLE OUTCOME
The first occurrence is retained. Every later occurrence of the same reference is flagged for review, with its reason. Uncertain decisions stay visible.
02distinct references
01repeated row flagged
0 bank movements · 0 real personal data
ENGAGEMENT PROCESS
Start with the problem and a genuinely anonymous sample. Agree written scope, price, acceptance criteria and exclusions before delivery. No unnecessary access to banking or production.
These cases have no demonstrated client outcomes. Not accounting or tax advice. No real personal data is accepted until appropriate safeguards are qualified.AEGIS / Examine
Choose your way in.
One body of work, four questions.
Local guide · prepared responses · no connected AI
Examine this record ↘How to read this record
SELF_REVIEW: reviewed within the same authorship chain. FAIL: a criterion fails within the test’s scope. BLOCKED: a prerequisite is missing before a test or decision. None implies independent validation.
The six records describe specific versions. F01 retains an interpretation dispute; F02 lacks the original vulnerable-run stdout. Built does not mean validated or production-ready.
Built.
And questioned.
Three things built. Three things the records forced us to question.
01—06Six records, sources and limits: evidence-en.html
01 / Built
02 / Failures and withdrawn conclusions
Guided exploration requires JavaScript. All six summaries remain available here.
A conflict test suite
Currentness v0.2 preserves 24 cases, code and results: latest-write-wins 10/24; authority-first 16/24; reference 24/24 against author labels. This is consistency within that corpus, not independent semantic validation. Later self-review found additional cases that challenge generalizing the resolver. See F01.
Original-source status · documentary review: Currentness v0.2 release package recovered with matching SHA-256 (1/2 required originals in this group). The original adjudication_results.json is missing; labels and semantic interpretation lack independent adjudication.
Open evidence excerptAn experiment lab
Experiment Foundry R2 has source code, evaluation records and a 17-scenario report. It is a synthetic laboratory. The R2 version reviewed here is recorded as R2_NOT_VALIDATED. It establishes neither real-world performance nor an independently qualified evaluator. See F02 for an R1 failure and the bounded R2 repair.
Original-source status · documentary review: Foundry R1/R2 original sources outstanding: 7/7 without matching SHA-256 in this review. Reported outcomes come from historical summaries and self-review; the experiment was not rerun here.
Open evidence excerptA source-reading prototype
TRACE V1.1 is a local, decoupled adapter with a prospective protocol. The retained suite reports 108/108 checks; the captured sample produced ABSTAIN. SELF_REVIEW, with partial source coverage. This is not native Core integration, predictive validation or evidence of economic utility. This card describes V1.1; later V1.2 work is separate.
Original-source status · documentary review: TRACE V1.1 original sources outstanding: 5/5 without matching SHA-256 in this review. The historical 108/108 figure is attributed to the version record; it was not rerun here and does not establish predictive usefulness.
Open evidence excerptA passing suite was insufficient
In the retained case, B corrects C and A remains unlinked. Output B is reproduced; UNKNOWN is the author’s expectation under local precedence. The 15 October review leaves open a rival global-priority interpretation and possible author misinterpretation. SELF_REVIEW / DISPUTED: a counterexample to the proposed criterion, not an unambiguous bug or independent adjudication. The original 24/24 does not settle this dispute.
Original-source status · documentary review: Currentness v0.2 release package recovered with matching SHA-256 (1/2 required originals in this group). The original adjudication_results.json is missing; the counterexample interpretation remains disputed and independently unadjudicated.
Open evidence excerptConsistent reports were insufficient
The retained summary of the historical review states that an authorized falsifier supplied 11 self-consistent reports for a false-causality scenario, that R1 accepted them and that the honest control failed. That summary attributes to R2 the rejection of the retained attack as INVALID_EXPERIMENT, with the holdout closed. The historical review reported inspecting the R2 receipt and the R1 audit, without rerunning R1 or recovering its vulnerable-run stdout. In this website review, those originals have not been recovered and Foundry has not been rerun; these claims remain pending verification. R2 remains unvalidated.
Original-source status · documentary review: Foundry R1/R2 original sources outstanding: 7/7 without matching SHA-256 in this review. The negative result is a summarized historical record, not a new execution or independent audit.
Open evidence excerptA receipt check could approve nothing
The retained R1 evaluation records PASS_BOUNDED, delivery_allowed=true and evidence_grade=NONE for null. Each of two runs found three false acceptances among 36 contract negatives. The package verdict is KILL. This is a failure of that candidate validator, not evidence that AEGIS performed an external action. SELF_REVIEW.
Original-source status · documentary review: Receipt Gate original sources outstanding: 3/3 without matching SHA-256 in this review. The anomaly is reported from a historical summary; it was neither reproduced nor independently verified here.
Open evidence excerptA QUESTION THAT REMAINS OPEN
If an institution outlives its founders, what should preserve its identity and limit their authority over its future?
AEGIS has not resolved this.
AEGIS / BOUNDARY
Authority
has a boundary.
You can examine this page. This page cannot act on AEGIS.
01
Selected records only
This page contains a fixed selection of evidence for review.
02
No delegated action
No access to the core, external tools, accounts or credentials.
03
A visible boundary
This version has no connected agent. Its guide offers prepared responses.
OPEN WORK
The next idea may come from someone else.
AEGIS is under construction. Contributions should be assessed by their evidence, limits and what they help us learn. An idea’s origin does not determine its value; allocating resources still requires priorities.
A missing reference, a small counterexample or a replication may change the direction. We propose agreeing on attribution, scope and terms before incorporating work. Writing to us does not transfer rights or grant system access.
Propose a contribution ↗CONTACT / CHOOSE YOUR PURPOSE
One conversation.
Two distinct purposes.
Choose research or professional work. Neither path connects to an automated agent.
Email hola@blancomendez.com or copy the address into your usual email service.
Questions, papers and collaboration
A counterexample, missing reference or substantive technical discussion.
Discuss a project
Describe the problem and expected outcome. Do not send real client personal data yet.
Do not include credentials or client personal data in an initial message. You can request privacy assistance through this mailbox.