Research orchestration
Declare the scope, coordinate the registered research sequence and reconcile the result.
Operating boundary
Human-invoked. The full series is optional; individual projects remain independently usable.
THE METHOD / JOHN COLLINS
Follow the mechanism.
Keep the missing link.
A claim about AI can be technically true and economically incomplete. My work follows the chain far enough to find out what a decision would require.
Try the public example ↗THE RESEARCH DISCIPLINE
Five habits.
One accountable judgment.
- 01
Start with the change.
Establish the source, date, identity and actual scope. An announcement, a deployment and a measured outcome are different facts.
- 02
Follow the mechanism.
Trace capability through constraints, spending, delivery and economics. Identify the first link that the evidence cannot support.
- 03
Try to break the inference.
Test timing, comparison groups, magnitude and rival explanations. Leave the unexplained part visible.
- 04
Make the work inspectable.
Publish the inputs, calculation, limitations and dated result. A configured process is not proof that it ran.
- 05
Let correction compound.
Preserve the earlier judgment. Record what changed and carry that correction into the next question.
THE WORKING SYSTEM
Eleven projects.
Different questions.
John Collins Alpha connects specialist research with a shared discipline for identity, evidence, scope and later observation. These are implemented roles, not eleven claims of proven efficacy.
Declare the question→Discover & verify→Challenge & record→Observe & revise
Live opportunity research
Investigate current market change and compare candidates inside a finite declared opportunity set.
Operating boundary
A current observation is not evidence of its cause. Account context enters only when needed.
Session reconstruction
Reconstruct the completed session, test explanations and translate supported changes into research judgments.
Operating boundary
Regular trading, after-hours and premarket observations remain distinct.
Catalyst calendar
Reconcile dated events, source documents and dependencies that can change a decision.
Operating boundary
Dedicated invocation, outside the Root series. Changes need provider read-back.
Primary filings
Ground company claims in issuer and regulatory disclosures, with dated evidence and identity checks.
Operating boundary
A filing's existence alone is not an investment signal.
Citadel Securities intelligence
Test public Citadel Securities market-intelligence, positioning and market-structure claims.
Operating boundary
A finite specialist source surface, not complete general-news coverage.
Specialist newsletters
Identify specialist claims and missing mechanisms, then seek independent corroboration.
Operating boundary
Paid text and private source material do not enter this public site.
SemiAnalysis research
Investigate AI architectures, physical constraints and commercial transmission beyond a source summary.
Operating boundary
Technical importance does not establish a listed beneficiary or shareholder return.
Portfolio stress
Test liquidity, margin, path dependency and survival under adverse conditions.
Operating boundary
Dedicated invocation, outside the Root series. A stress scenario is not a forecast.
Public work & memory
Make supported work, reasoning and corrections inspectable through Deep Vertical AI and John's public profiles.
Operating boundary
Private evidence informs investigation; only reviewed public-safe material is published.
Recommendation evaluation
Append later observations to immutable recommendation boards and their exact opportunity sets.
Operating boundary
Scores project recommendations, not human decisions. Interim observations are not terminal efficacy.
The registry-defined Root series currently combines Projects 01, 02, 04, 05, 06 and 07, then Project 10 evaluation. Projects 03, 08 and this public surface have their own invocation. Runs are human-initiated; a diagram is not evidence that every role ran at a particular cutoff.
BUILT AROUND THE RESEARCH
The infrastructure
behind the answer.
Scope & identity
Declared finite opportunity sets, exact security identities, source cutoffs, versioned contracts and schema validation. No claim to a complete global AI-stock universe.
Evidence & artifacts
Source provenance, calculations, workbook checks, rendered review and independent challenges. Conclusions are qualified at the first unsupported step.
Memory & evaluation
Immutable boards are paired with their exact opportunity sets. Later observations append to the record. Investment and recommendation evaluation use exactly 0–3 months; interim observations stay separate from terminal scores.
Publication & control
Versioned public editions, dated corrections, protected contact, candidate deployments and verified rollback. Human authority governs consequential decisions; research access does not become trading authority.
IMPLEMENTATION ≠ VALIDATION
The complete continuous memory layer is unfinished. Published recommendation process evidence does not establish a profitable track record. The public site exposes reviewed work, not the underlying private account, source corpus or operational records.
Inspect the evaluation boundary ↗PUT THE METHOD TO WORK
What would you need
to know to decide?
Start with one consequential research question and a clear evidence boundary.
Discuss the problem ↗