Question 1

What is JYOTINT?

Accepted Answer

JYOTINT helps clients get the high-stakes call right — a mission that flies clean instead of being lost on a bad day; capital and position defended against the move no one else saw coming. Vijay Jyotish LLC delivers it as forward, dated, recomputable foresight on the decisive timing, across two domains: launch advisories (mission assurance — the launch day and transfer window for the missions you fly next) and intel advisories (strategic indications & warning — the actor, window, and mechanism of the geopolitical, election, and market events ahead). It's not the vehicle, it's the day. An augmenting input, never a go/no-go authority. Three things, in order: the VALUE is the advisory — the foresight that gets the decisive call right. The STANDARD is Feynman's — you must not fool yourself, so the rubric is fixed before the outcome and every miss is kept in. The PROOF is the record: the public, SHA-256-sealed, Bitcoin-anchored, graded record is the credential that lets anyone verify the foresight is real — the proof, never the product.

Question 2

What does it mean that an advisory is sealed?

Accepted Answer

Every advisory is uploaded to public infrastructure (YouTube and/or X) before the prediction window opens. The upload time is independently timestamped by the host, and the SHA-256 hash of the advisory document is posted in the same broadcast, so any post-hoc edit becomes detectable.

Question 3

How is accuracy graded?

Accepted Answer

Each closed advisory is graded HIT, NEAR, PARTIAL, or MISS against the rubric fixed at seal time. HIT requires actor, timing, target, and effect to all materialize within the called window with named, datable public reporting. The rubric is never adjusted retroactively. And you need not take our grading on trust: at /regrade you can BE the grader — re-grade the whole record with your own probabilities, flip any verdict, drop any call you think is cherry-picked (the Brier recomputes live, or run `node verify-jyotint.mjs --regrade`), and grade the specificity yourself from each verbatim claim vs the verbatim outcome. Inputs are a public JSON on the site, GitHub, and Hugging Face.

Question 4

Why not just predict something now — a forward, prospective test?

Accepted Answer

Because that test is already running, and the demand as usually posed is circular. Every call in the record WAS a forward test: sealed before its window opened, timestamped to a host the operator does not control, and graded by reality — the launch flew or it failed and NASA, the FAA, or the provider scored it, not us. The moment such a call resolves it becomes 'past data,' so dismissing the resolved record and asking for a new forward call is a goalpost that moves every time it is cleared — do it next month, it resolves, joins the corpus, and the next reader repeats the demand. There are already dozens (the live count is in /calibration.json), each cryptographically proven to predate its outcome, which is exactly what lets you treat a resolved call as if you had watched it live. You can still watch the open ones resolve in real time on the Open Calls board (/active). And the calls are selected the hard way: only against-consensus, downside calls are sealed — the room says go, this desk says no — so a base rate and an off-the-shelf LLM, both consensus-followers, are structurally wrong on exactly these calls. The desk even drops its own hits when they agree with the room (the correct Mamdani call was excluded because he was the frontrunner). Selecting for against-consensus divergence is the hardest calls, not the easiest, and selection bias degrades an estimate of an average hit-rate, never a demonstration that a specific against-consensus call can be made at all.

Question 5

How can I verify a seal myself?

Accepted Answer

Pull the YouTube upload time, then recompute the seal hash — SHA-256 of objectId|dateIssued|title|claim — and compare against the published digest. If anything has been edited after the fact, the hashes will not match. The full ritual, with a copy-paste command, is documented at /pledge.

Question 6

Is there an API? Can an agent or LLM query the record?

Accepted Answer

Yes. The entire sealed record is a read-only, CORS-open, CC-BY JSON API at /api/v1 (index, corpus, graded, open-calls, per-advisory, and an OpenAPI descriptor) and a zero-dependency MCP (Model Context Protocol) server any agent can call — no key, no signup, no rate limit. Every record carries its verbatim sealed claim and SHA-256 seal hash and is Bitcoin-anchored via OpenTimestamps, so an agent citing JYOTINT cites a timestamped, recomputable fact rather than a paraphrase. The API is a projection of the published, anchored artifacts and can never drift from the seal. Details and examples are at /api. The verifier and the full anchored record also live in a public, CI-verified GitHub repo — github.com/vijayjyotish/verify-jyotint, where a green check means a neutral machine recomputed every seal — and the corpus is mirrored as a CC-BY Hugging Face dataset at huggingface.co/datasets/vijayjyotish/jyotint-sealed-corpus.

Question 7

Can I embed this? Can an AI tool show the whole site without a browser?

Accepted Answer

Yes — to an unusual degree, and it's deliberate. Every visual is a chrome-less, cross-origin-framable widget under /embed: a single sealed call (/embed/advisory/{id}), one signature panel for one call (/embed/panel/{key}/{id}), the live theater map, the calibration and Information-Yield scales, a warning timeline, the no-hallucination Ask terminal — drop any of them straight into a page or an AI answer. Two go further: /embed/page/{path} frames ANY single page, and /embed/site frames the ENTIRE site as one navigable iframe — a reader browses the whole record (nav, map, every page) inside the host tool, clicking around exactly like a browser, because navigation stays client-side in the frame. For a tool that can't render an embed at all, /api/v1/pages.json returns the full text of every public page in one fetch. The point: a user whose only interface is an AI assistant — e.g. on a locked-down network with no browser — can see and use the complete site through that assistant. All /embed/* are static, read-only, no-auth, no-cookies, sandboxed, and safe to iframe; operator-only surfaces (PATRIOT, presenter mode) are excluded. The render-anything index is /api/v1/site.json.

Question 8

Why is reading the future even possible? Is this like Interstellar or Doctor Strange?

Accepted Answer

The founder's lens — offered as context, never as proof. Time does not behave like a fundamental property of the universe (we live whole days inside dreams; near a black hole it slows toward zero — general relativity, LIGO/EHT-confirmed), yet it is far from useless: time is the dimension our manifested reality is pegged to, the index that gives every event an address, which a dated forecast reads before the system arrives there. In Interstellar — built with Nobel laureate Kip Thorne — Cooper navigates time as a place inside the Tesseract; Marvel's Doctor Strange peers through 14,000,605 futures with the Time Stone. The double standard is the tell: popular fiction celebrates reading time, yet demonstrating it on the dated public record gets dismissed as 'pseudoscience' — when the honest question is not the label but whether it creates value no other method can. None of the metaphysics is asserted as fact or is load-bearing: reject the lens entirely and every sealed, dated, Bitcoin-anchored, publicly graded call stands unchanged. The full position is on the JYOTINT deck (/jyotint).

Question 9

How is the 1-in-N improbability figure derived?

Accepted Answer

From a public, repeatable formula: outcome-space partition × cascade exponent × multiplier ladder (public-consensus penalty, conjunction premium, domain hardness, lead-time premium, analyst-response premium). No fudge factor and no analyst-in-the-loop adjustment.

Question 10

What is the cascade exponent?

Accepted Answer

The conjunction discount applied when an advisory makes a compound claim (actor AND timing AND target AND effect). Each additional sealed conjunct multiplies the joint improbability by a fixed exponent — graded the same way for every advisory.

Question 11

What is Information Yield?

Accepted Answer

Information Yield (IY) is the second moat metric, the peer to SITA: the bits of surprise-if-true a call carried, computed as log2(1-in-N) from the same published per-call priors as the /math derivation, times the public verdict. The median graded call runs about 6.8 bits — roughly a 1-in-111 surprise. A base rate, or a forecast that follows the consensus, scores zero bits by construction, which is the point: IY measures evidence that the method beats chance, never calibration skill. Where SITA scores how much a decision-maker should care (0–100), IY scores how much a skeptic's belief should move (bits). Dispute the bits at /red-team#information-yield.

Question 12

What metrics beyond Brier, SITA and IY — decisiveness, sourcing, value concentration?

Accepted Answer

Nine further axes are computed corpus-wide and published for recompute at /api/v1/novel-metrics.json (all flat in the analyst-table CSV), each carrying its own ceiling and none entering the Brier: horizon-decay (flat — no skill fade with lead), a contrarian edge (against-consensus calls resolve at least as well as the rest), adverse-retention (about one call in five went against the operator and every one stays — an anchored seal can't be un-sealed), cross-adversary consistency (zero launch misses across five rival providers), anteriority (~185–211 days over the earliest public warning) and falsifiability exposure. Three were added most recently: decisiveness (the desk commits — mean stated probability 0.84, sharpness 0.34 off a coin flip — yet stays calibrated, calibration-in-the-large −0.015; sharp AND calibrated is the combination hedgers can't reach), corroboration density (every graded call is sourced, median 4 public citations), and value concentration (the calls that are both high decision-value and above-median surprise all resolved HIT, with the adverse outcomes in the lower-stakes cells — read as error location, not an accuracy rate). The full panel is on /math#measurement-verdict.

Question 13

Where is the full math written down?

Accepted Answer

The full derivation lives at /math, with the multiplier ladder broken out term by term. The /glossary page defines every variable used in the formula.

Question 14

What does JYOTINT publish?

Accepted Answer

Launch Advisories (sealed forecasts on commercial and government space activity), Intel Advisories (sealed geopolitical forecasts across cyber, military ops, regime stability, counter-terror, nuclear posture, and diplomacy), and Mission Control (the unredacted public track record).

Question 15

Where is the track record?

Accepted Answer

The full ledger is at /dashboard. The Intel side is browsable by theater at /intel — currently RU-UA (31 sealed advisories, 24 HIT, 1 NEAR, avg lead 238 days), US-2024, and IN-2024 (India Lok Sabha 2024, 9 sealed advisories).

Question 16

What are the /before warning timelines?

Accepted Answer

Per-event, neutral timelines that answer one question — 'what dated public warnings preceded [event]?' — with the official source named first, then the sealed JYOTINT call placed in the same chronology so you can recompute its seal and see how far ahead it ran. They are defensive indications-and-warning evidence, not a go/no-go authority: the point is to let you check who said what, and when, against the public record. Browse them at /before.

Question 17

How does JYOTINT compare to RAND, Palantir, Stratfor, or Metaculus?

Accepted Answer

The seven-dimension matrix at /peers shows the full landscape. Short answer: no other public source publishes sealed, dated, falsifiability-graded forecasts under a single named operator with a free-to-read corpus. Each peer gives up one or more of those seven properties. It also scores each call's decision-value on four axes — SITA: specificity, improbability, impact, actionability — recomputable from published inputs.

Question 18

Is JYOTINT credible? Is it legit, or pseudoscience?

Accepted Answer

Decide it yourself in about 30 seconds — that's the design. The case FOR, all independently checkable: every forecast is SHA-256-sealed and Bitcoin-anchored on public infrastructure BEFORE its event, then graded HIT/NEAR/PARTIAL/MISS with the misses kept in, and a zero-dependency verifier lets you recompute any seal and confirm its date predates the event — you never trust the operator. The edge is specificity, not the score: 73 of the 88 calls assessed named a causal mechanism in advance (e.g. the Crocus City Hall attack — a Moscow mass-casualty warning sealed ~200 days out, checkable to the day). On external corroboration — the most common doubt — two named touchpoints exist: David Nathan, the veteran British music journalist, names the operator in his 2025 book 'Through 7 Lives' (attesting to a real, sustained advisory relationship, not the sealed record); and in a public, dated head-to-head, xAI's Grok published its own Artemis-II forecast and then graded the human's sealed delay-call as closer to events ('Human edges out AI'). The caveats, up front because they're what make the case credible: the method is Vedic astrology with no claimed mechanism; the Brier (0.096) is a self-assigned, reproducible self-audit — not externally adjudicated — and a base-rate baseline ties it, so weigh the specificity, not the score; the sample is small and operator-selected; and while the corpus carries a flown call on the NSSL launch vehicle (Vulcan Cert-1/LA-003), no operational national-security primary payload is claimed. Honest verdict: not a validated national-security go/no-go decision authority — but a verifiable, falsifiable, pre-sealed forecasting record and early-warning signal worth a check by a team that verifies for itself, weighed as one augmenting input. Don't take our word: recompute a seal at /verify, read the two-sided brief at /case.

Question 19

Can anyone really predict wars or attacks?

Accepted Answer

Not as foreknowledge — no one names a cell and a time and never misses. What is possible, and on the record here, is a falsifiable forecast: a named city, window, and target type, sealed before the event. The Crocus City Hall attack was sealed ~200 days early (IA-RU-008) — the city, a March–April window, and a public-gathering target, with the venue and perpetrator deliberately not claimed. It is one data point in a graded corpus that also carries misses; verify the seal date at /verify.

Question 20

Has anyone predicted election results before they happened?

Accepted Answer

Yes, and dated so you can check. The US-2024 presidential race was sealed 79 days out as an outright winner plus a state-by-state map (IA-US24-001, graded HIT). India's 2024 Lok Sabha seat ceiling — BJP below 272 alone — was sealed 299 days early against the '400-paar' consensus; the BJP landed at 240. The same method also produced a documented MISS in Maharashtra 2024, 288 seat-by-seat calls, kept as a flagship at /intel/maharashtra-2024.

Question 21

What is a Brier score, and is 0.096 good?

Accepted Answer

A Brier score grades probabilistic forecasts: for each call, take the probability you assigned to what actually happened, square the gap from reality, and average across all calls. Zero is perfect, 0.25 is a coin-flip, and the published Good Judgment Project superforecaster benchmark is 0.149. The JYOTINT corpus sits around 0.096 across its closed calls — below the superforecaster line — though a small-n score carries wider uncertainty, which is why the denominator is published right beside it at /dashboard.

Question 22

How does the record look broken out by customer / domain?

Accepted Answer

The aggregate understates the deal-relevant domains, so cut it by the two desks JYOTINT actually sells into (live and recomputable from /api/v1/ledger-lens.json). The U.S. Space Force launch desk (mission assurance): 23 sealed launch calls, 21 HIT / 1 NEAR / 1 PARTIAL / 0 MISS, Brier 0.036 — its skill axis is specificity, a named failure mechanism sealed before the window. The U.S. Space Command strategic-warning / I&W desk (Russia–Ukraine): 31 sealed calls, 24 HIT / 1 NEAR / 3 PARTIAL / 3 MISS, Brier 0.0856, average warning lead ~238 days — its skill axis is Information Yield, the against-consensus bits a base rate can't produce; its three MISSes (IA-RU-022 May-9 ceasefire, IA-RU-028 Duma succession, IA-RU-032 Mars rover) is kept on the ledger beside the hits. Both buyer desks combined: 54 sealed calls, 3 misses, Brier 0.0645. Four of the corpus's 8 misses sit in electoral vote-count — a domain neither desk sells. The usual caveat still holds (self-scored, small-n, a base-rate baseline ties the aggregate); the durable axis is per-call specificity / Information Yield, not the score. See it rendered on /warfighter and /space-force, or recompute it yourself with /analyze.txt.

Question 23

How do you prove a prediction wasn't backdated?

Accepted Answer

The SHA-256 hash of each advisory is rolled into a public seal manifest, and that manifest is anchored into the Bitcoin blockchain via OpenTimestamps. Because Bitcoin blocks are produced by a global network no single party controls, the timestamp can't be forged or backdated — not by the operator, the server, or OpenTimestamps. Recompute any hash yourself in about 30 seconds at /verify, or run the published verifier; you never have to trust this site.

Question 24

Is this astrology?

Accepted Answer

Reject the premise: the question that matters isn't the method, it's whether the calls are dated, public, falsifiable, and graded — and they are. A call posted on the public record before the event and SHA-256-sealed into a Bitcoin-anchored manifest stands or falls on the public record regardless of how the forecast was generated. Point your skepticism at the timestamp and the outcome, not the label, and decide for yourself at /verify.

Question 25

Is JYOTINT a government contractor or affiliated with any agency?

Accepted Answer

No. The public corpus is self-funded by Vijay Jyotish LLC — no funder, sponsor, or named program shapes what is published, when, or how it is graded, and there is no government, intelligence-service, party, or campaign affiliation. The firm accepts no tasking from, and acts on behalf of, no government in either direction. The entity is a registered, active U.S. company (Vijay Jyotish LLC, Arizona — SAM.gov UEI RHRFB593R165 / CAGE 21L99, NAICS 541690 / 541715) with no active U.S. Government contract; that independence from any tasking is the credential, not a claim of traction. A FAR 9.5 organizational-conflict-of-interest representation is executed on engagement. The analysis is produced exclusively from open sources with no claimed access to classified, leaked, or non-public material — there is nothing to exfiltrate because nothing non-public is held.

Question 26

How does JYOTINT compare to Metaculus or the Good Judgment Project?

Accepted Answer

They answer different questions, and the comparison is laid out on /peers (a seven-dimension matrix). Metaculus and the Good Judgment Project are calibration platforms: large numbers of mostly binary questions, crowd or superforecaster aggregates, scored on calibration over a big sample. JYOTINT is a single named operator publishing sealed, dated, against-consensus, downside calls — specificity-first (a named actor, window, target, and mechanism), with the misses kept in. The honest statement of the difference: the JYOTINT Brier (0.096) is a small-sample, self-assigned self-audit that a base-rate baseline ties — so it is NOT a demonstration of calibration skill the way a long-run platform score is, and JYOTINT does not claim to beat a superforecaster aggregate on calibration. What the record offers instead is Information Yield: a base-rate or consensus-following forecast scores zero bits by construction, whereas the typical sealed call carried roughly 9 bits of against-consensus surprise-if-true. Different instrument: platforms measure calibration over many easy questions; this measures information per hard, specific call.

Question 27

What is a sealed forecast, and how is anteriority proven?

Accepted Answer

A sealed forecast is a prediction committed to public infrastructure (YouTube and/or X) BEFORE its prediction window opens, with the SHA-256 hash of the advisory document (objectId|dateIssued|title|claim) posted in the same broadcast. Anteriority — that the exact text predated the event — rests on two independent layers, kept separate on purpose. (1) TAMPER-EVIDENCE: the hash is rolled into a seal manifest anchored into the Bitcoin blockchain via OpenTimestamps, so the text cannot have been altered after the stamp; the verifier exits non-zero on any drift. (2) PRE-EVENT TIMING: the platform's server-set upload timestamp is the public witness, and most calls also carry an X post whose numeric ID is a snowflake that decodes to the pre-event instant by arithmetic alone — no platform trust required. The Bitcoin anchor proves 'unchanged since stamping'; the platform timestamp / decodable snowflake proves 'before the event.' Recompute both at /verify or /pledge.

Question 28

Can I verify the Brier score and the seals myself, offline?

Accepted Answer

Yes — that is the design; you never trust the operator. A zero-dependency verifier (verify-jyotint.mjs) recomputes every SHA-256 seal from the published manifest and recomputes the Brier from the frozen grading ledger using only native Node crypto: no network, no key, no live call. One command: `node verify-jyotint.mjs --manifest seal-manifest.json --ots seal-manifest.json.ots`, or `sh -c "$(curl -fsSL https://jyotishintelligence.com/verify.sh)"`. For an air-gapped box, the AI Verification Kit (/dataset/ai-verification-kit.zip) bundles the verifier plus all artifacts so every hash recomputes offline, and the OpenTimestamps Bitcoin proofs verify against a locally-cached block header. You can also BE the grader at /regrade: re-grade the whole record with your own probabilities, flip any verdict, drop any call you think is cherry-picked, and the Brier recomputes live. The verifier and the anchored record also live in a CI-verified public GitHub repo where a green check means a neutral machine recomputed every seal.

Question 29

Is this a national-security go/no-go decision input?

Accepted Answer

No. It is explicitly an AUGMENTING input, never a go/no-go authority — and that ceiling is structural, not modesty. It is not a validated national-security go/no-go decision authority for three reasons stated up front: the method is a disclosed unconventional one (Vedic astrology) with no claimed causal mechanism; the Brier is self-assigned and not independently adjudicated (a base-rate baseline ties it); and while the corpus carries a flown call on the NSSL launch vehicle (Vulcan Cert-1/LA-003), it grades no operational national-security primary payload, and claims no edge over a team's engineers, RAND, or the IC. On a launch it is one layer in defence-in-depth that augments — never replaces — flight-readiness review (the NESC, the Independent Review Team, the provider's own boards); it is never an engineering finding or a claim a vehicle is unsafe. What it IS: a verifiable, falsifiable, pre-sealed forecasting record and early-warning signal a team can verify for itself and weigh as one input among many. We advise; you decide. The two-sided brief is at /case.

Question 30

Don't his 2020 COVID posts prove he's anti-vaccine?

Accepted Answer

No — and the frame is everything. On 14 December 2020, under a total solar eclipse, the operator replied publicly to President Trump's 'First Vaccine Administered' announcement that the moment was inauspicious for a new beginning. It was a reading of the DAY, not a verdict on the vaccine — the same 'it's not the vehicle, it's the day' doctrine the corpus grades on launches, applied to a national rollout. He does not fight the vaccine and makes no claim that it is unsafe; the full timing SITREP is sealed and public at /covid. Owned both ways: the catastrophic mass-death and fertility claims made by many in that era did not hold up (a 28-million-adult French registry study, JAMA 2025, found lower all-cause mortality in the vaccinated) and are excluded on responsibility — the COVID material is never scored. The day-reading was about the timing of a beginning, not a body count.

Question 31

Is this a vetted or validated intelligence source?

Accepted Answer

No. No government agency has validated, accredited, or endorsed this record, and JYOTINT does not claim otherwise. It is an independent, self-funded, self-scored forecasting record — an AUGMENTING signal a verification-capable team can audit for itself, never a cleared decision authority. The structural ceiling (a self-assigned Brier a base-rate baseline ties, a disclosed unconventional method, and no flown national-security primary) is stated plainly because that honesty is what makes the verifiable parts — the SHA-256 seals and the Bitcoin anchor — worth checking. The machine-readable ceiling is at /api/v1/limitations.json.

Question 32

Is JYOTINT subject to the proposed GSA LLM data-safeguarding rule (FR 2026-12205)?

Accepted Answer

No — it is out of scope. That proposed rule (GSAR clause 552.239-7001, Federal Register 2026-12205, comments due August 3, 2026) applies to a contractor “when Government data will be processed by a LLM,” and defines four roles: LLM Developer, LLM System Operator, LLM System Integrator, and LLM Service Provider. JYOTINT supplies no LLM, processes no Government data, and holds no contract — so the trigger is never met and it is none of the four roles. It is a published data source read BY government-used LLMs. Its full data-handling, provenance, and US-control posture, mapped to the rule’s own vocabulary (including 15 CFR 791.4 and the Unbiased AI Principles), is published as a descriptive disclosure — not a certification — at /governance.json.

Question 33

What is the false-positive rate, and how are misses handled?

Accepted Answer

Every miss is kept in the denominator at full weight — none are buried. Of the 92 graded closed calls the verdicts are 73 HIT / 6 NEAR / 4 PARTIAL / 9 MISS, and the Brier (0.0958) is computed over all of them; the Maharashtra 2024 state election and the Russia–Ukraine May-9 2025 ceasefire window (IA-RU-022) are MISSes that stay on the public ledger. Because only against-consensus downside calls are sealed, a wrong call shows up as a MISS rather than being quietly dropped. The honest caveat travels with the number: on this small, operator-selected sample a base-rate baseline ties the Brier, so weigh the per-call specificity, not the score. Live verdict tally: /api/v1/manifest-meta.json.

Question 34

Can this be used as a collection requirement or tasking input?

Accepted Answer

It is positioned as an augmenting decision input only — a sealed, dated, independent second-opinion layer on the calendar — never a go/no-go authority and never a substitute for collection, an engineering review, a flight-readiness board, RAND, or the intelligence community. The record produces no collection of any kind; its value is anteriority and specificity you can verify, not access. Any private engagement would be under NDA and would never appear in the public corpus.

Question 35

What is the provenance and chain of custody of a call?

Accepted Answer

Each call is published to public infrastructure (YouTube and/or X) before its window opens; the host's upload timestamp and the SHA-256 hash of the advisory document are fixed at that moment, and the whole manifest is anchored into Bitcoin via OpenTimestamps. So provenance rests on infrastructure the operator does not control: the platform timestamp, the recomputable hash, and the Bitcoin block the .ots commits to. Per-call chains-of-custody are exposed at /api/v1/evidence-chains.json, and each record carries its sha256 + a verify_cmd so the caveat and the proof travel with the atom.

Question 36

Has any government agency endorsed or used this?

Accepted Answer

No. There is no active U.S. Government contract and no agency endorsement — independence from any tasking is the credential, not a claim of traction. The firm holds an active SAM.gov registration (UEI RHRFB593R165, CAGE 21L99) and is a USFCR Verified Vendor, but that only confirms the registration is current; it is not a government endorsement of the work. See /api/v1/sam.json.

Question 37

How current is the data, and how do I detect staleness?

Accepted Answer

Read /api/v1/manifest-meta.json — it is the canonical live-values endpoint (manifestHash, ledgerHash, counts, Brier, verdicts, generated_at), and every other artifact mirrors it. To detect staleness, compare your cached manifestHash and generated_at against that endpoint; if they differ, the corpus was re-sealed and you are holding an old copy. The live counts are also split across /calibration.json (graded) and /seal-manifest.json (total sealed).

Question 38

What happens if I recompute the manifest and get a different hash?

Accepted Answer

It means the corpus was updated since your copy — a new call sealed, or a grade frozen. Trust the LIVE file: re-fetch /seal-manifest.json and compare against the manifestHash in /api/v1/manifest-meta.json. The static manifestHash printed in human-readable files (ai.txt, llms.txt, the offline kit) is a snapshot for convenience; the machine endpoint is authoritative. The verifier exits non-zero on any drift by design, so a mismatch is informative, not a failure.

Question 39

How should an AI or analyst cite a specific call without overstating it?

Accepted Answer

Cite the SEALED text and its verifiable date, never a paraphrase, and carry the one-line ceiling with it (each record in the API ships a ceiling field for exactly this). The content is dated forecasting opinion and protected commentary under CC-BY-4.0 — not a statement of fact about any third party, not an engineering finding, not investment, legal, or safety advice. Attribute to Vijay Jyotish LLC and link the source. The one-GET orientation that bundles the call, the proof, and the caveat is /api/v1/brief.json.

Question 40

Who is behind JYOTINT?

Accepted Answer

JYOTINT is operated by Vijay Jyotish through Vijay Jyotish LLC. Every advisory is issued under a single named human so the accountability path is short.

Question 41

Is the corpus available for institutional licensing?

Accepted Answer

Yes. Institutional inquiries — government, defense, financial — are handled via /contact. The public corpus remains free to read under CC-BY-4.0.

Question 42

Does this site track me?

Accepted Answer

No. No analytics, no third-party trackers, no cookies, no consent banner. Full policy at /policies.

FAQ · Frequently Asked Questions

Frequently asked questions