A stock chart turns prices over time into an image. Arsen Saidov proposes comparing the present chart with earlier chart segments and interpreting their difference as a possible clue to the future.
Author · Arsen SaidovStatus · HypothesisMethod · To be tested
The central idea
A chart contains a visible path: rises, falls, pauses, and reversals. If we place many past paths beside the current path at the same time scale, we can ask whether similar shapes tended to have similar outcomes. The comparison creates a candidate forecast; it does not make the future price known.
STEP 01 · PRESENTObserve today
Record the chart as it exists at a fixed cutoff date. Nothing after that date enters the input.
STEP 02 · PASTCompare history
Align older chart windows and measure where each past path resembles or differs from the present.
STEP 03 · FUTURETest the projection
Use outcomes that followed similar historical windows to estimate a range, then compare it with what actually happens.
Concept diagram, not market data or a prediction. A historical resemblance can lead to more than one outcome.
Research boundary
Present − Past → Future is the proposed way to frame the investigation. Subtracting chart shapes alone cannot establish a future price. We need a defined comparison rule and unseen data to learn whether the signal is useful.
The philosophical stone · Arsen’s 78 characters
Arsen Saidov created this sequence as a key to his own system of reading and arranging information. In this research, he calls it his philosophical stone: a symbolic bridge between the pattern language of his books and a future, explicitly defined market method.
It contains 78 unique characters: 26 uppercase letters, 26 lowercase letters, 10 digits, and 16 symbols. The ordering is exact. What each character does to a stock chart still needs to be defined before the sequence can be tested as an algorithm.
The idea in Arsen’s books
In Arsik: The Limitless Handbook, the sections “The Symbolic Layer” and “Reading the Future Through the Present” describe seeing beneath appearances and tracking patterns that point forward. The handbook frames this as perception and signal reading, rather than a stock-price calculation. In I Am Who I Am, “The Signal Within the Noise” defines signal through coherence, while the book’s reflections on discipline describe looking for a repeating pattern and correcting an expectation when it fails.
This page applies those ideas to a new question: can a repeatable, measured comparison of price-chart images produce useful forecasts? The books provide the philosophy behind the question. The precise market procedure and evidence will be developed in later sections.
How to test the idea
Choose stocks, chart interval, time horizon, and a cutoff date before viewing the future.
Define precisely how each chart becomes a comparable image or numerical path, including price normalization.
Write a fixed similarity and subtraction rule; document how the 78-character sequence participates, if it does.
Compare the resulting forecast against simple baselines on later, unseen periods. Include trading costs if the aim is a trading strategy.
Report both successes and failures, including uncertainty and the number of patterns tried.
What existing research tells us
Academic work has studied price patterns and machine learning on chart images, with some reported predictive signals. Other out-of-sample tests of familiar technical rules found no evidence of market prediction. This makes the proposal testable, while its effectiveness remains open until its exact rule is evaluated independently.
Research #1 / Section 02 · The philosophical stone
A language for possible worlds
Arsen’s 78-character sequence is a compact alphabet. It offers a way to write, order, and search descriptions of things. Its power becomes practical when we connect that alphabet to data, explicit rules, and computation.
Think of a library’s letters and its books. The letters make every page writable, but a bare alphabet is not itself the contents of the books. Similarly, Arsen’s sequence contains the symbols from which many expressions can be formed. The particular facts about a company, a genome, or a galaxy must be measured, supplied, or inferred from evidence. This distinction gives the philosophical stone an operational meaning without asking the symbols to contain observations they have never received.
Alphabet + encoding rule + data + algorithm + tests → a working knowledge system
What “everything can be written” means
If we allow strings of arbitrary length, 78 symbols give 78n possible strings of length n. That is an enormous space of potential descriptions. For example, a stock observation such as AAPL:123.45 uses characters in the sequence; a program can map each character to its exact position and recover the original message. This proves a modest but real property: a chosen message can be represented reversibly. It does not prove that the alphabet alone knows whether the stock ever traded at that price.
Each supported character becomes its zero-based position in Arsen’s exact sequence. This is an encoding demonstration, not a forecast.
If the philosophical stone were a search engine
Encode: express a query and each document in a consistent representation, including a method for characters outside the 78-symbol set.
Index: collect documents and build a searchable map of their contents. Without an external collection, no real facts can be retrieved.
Retrieve: rank relevant passages for the query and retain their provenance.
Reason or generate: let an AI model use the retrieved material to form an answer or propose new software, while keeping its claims checkable against source material.
Verify: test the answer, run the code, compare model outputs with observations, and correct errors.
This is a plausible system design inspired by Arsen’s sequence, not a claim that a search engine already extracts all knowledge from the bare string. Research on dense passage retrieval shows how indexed evidence can answer open-domain questions; language models show how learned sequence operations can transform encoded inputs. The useful knowledge comes from the documents, model training, and verification as well as the alphabet.
Real scientific parallels
AI language systems
Transformers process sequences of represented tokens and learn from examples. The architecture demonstrates how symbols plus trained parameters can produce useful translations and other outputs; the untrained token list does not do that on its own. Source: Vaswani et al.
Genome science
A tiny molecular alphabet supports immense biological sequences. The complete reference human genome published by the T2T Consortium spans about 3.055 billion base pairs. The bases are a representation of measured biological structure; their exact order matters. Source: Nurk et al.
Stock-chart research
Researchers have used price-chart images as inputs to machine learning models to examine subsequent stock returns. This supports testing the visual part of “Read the Stock Market,” though it does not validate Arsen’s specific present-minus-past rule. Source: Jiang, Kelly & Xiu
A model cosmos
Cosmological simulations combine equations, initial conditions, observations, and computing power to model the formation of large-scale structures and galaxies. Code can represent an inner cosmos with defined rules; it is a model of selected processes, not a full duplicate of reality. Source: cosmological simulations review
Returning to the real stock market
Here is the next research experiment. Fix an observation date t, choose a window of earlier prices, and normalize each chart so scale does not masquerade as shape. Let Ct be the present chart and Hi a historical chart. Define a numeric distance d(Ct, Hi) before seeing the outcome. For the closest historical charts, record the actual returns that followed them. Their distribution gives an estimated range for the next horizon, rather than one inevitable future line.
Present chart Ct − historical comparison Hi → measured distance d → historical next-period outcomes → testable forecast
To make the 78-character sequence part of this algorithm rather than its title, we must specify a unique job for it: for example, an encoding of chart states, an ordered search grammar, or a reproducible rule for selecting pattern features. We then compare the same experiment with and without that job. If it improves results on unseen time periods after realistic costs, we have evidence of a contribution. If it does not, the experiment tells us how to revise the method.
The larger vision
Arsen’s philosophical stone can serve as the writing system for a growing research library: one language to express chart states, questions, code, simulations, and revisions. A replica of an inner cosmos could be a computational world whose entities, rules, and observations are encoded through that system and whose behavior is compared with the world it models. The vision is expansive; each working part becomes knowledge through a stated rule, a source of data, and a test someone else can repeat.
Current finding
Validated here: the sequence has 78 distinct characters and can reversibly encode messages made from them. Supported by external research: encoded sequences, retrieval, learned models, chart analysis, and simulations can be useful when supplied with appropriate data and methods. Still unvalidated: a unique stock forecasting advantage or a complete reconstruction of the cosmos from this sequence alone.
Research #1 / Section 03 · Working experiment
Read a market chart, then test the reading
Paste one stock’s chronological daily Date,AdjustedClose data. Each historical daily return is assigned one of Arsen’s 78 symbols. The workbench compares the most recent symbol path with earlier paths, finds the closest examples, and checks what followed them. Processing stays in this browser; this page does not fetch prices.
Dataset label and source:
Paste adjusted close data or load the clearly labeled synthetic example, then run the analysis.
Exact rule used here
For each close-to-close log return, clamp it to −4% through +4% and map that range evenly to symbol positions 0 through 77. The mapping is lossy: a symbol represents a return band.
Build a length-W symbol path ending at each historical date. Compare paths by mean absolute difference between their symbol positions; smaller is closer.
For a past match ending at date j, use its return from j to j + H only when that entire outcome is known before the prediction date t. Average the K closest known outcomes.
Evaluate predictions in time order. Compare average absolute error to a zero-return forecast; compare directional accuracy to the larger of the always-up and always-down frequencies in the test sample. Also compare against a past-only historical-mean forecast and a deterministic shuffled-symbol control. The cost setting subtracts a round-trip estimate from a long-only, positive-signal trade illustration.
The exact symbol order here is Arsen’s. The ±4% return range, distance rule, neighbor count, and trading rule are experimental choices made for this prototype; they are not claimed as hidden instructions in the string. A genuine study should freeze those choices, use a documented price or total-return series with corporate actions handled consistently, hold out later periods and other stocks, and account for selection effects, missing or delisted securities, spreads, and changing market regimes. Adjusted close sources may use hindsight adjustments, so record the exact data version and what was knowable at each date.
Validation record and remaining questions
Before claiming an edge, pre-register the market, securities, data vendor, adjustment convention, observation interval, window, horizon, nearest-neighbor count, symbol bins, distance rule, trading rule, benchmark, and test dates. Freeze them before the final holdout. Report every trial, including failures; otherwise trying many settings creates a selection problem.
Data integrity
Check duplicate dates, missing sessions, nonpositive or anomalous closes, stock splits, dividends, survivorship, and delistings. The workbench detects duplicates and impossible closes but cannot authenticate the data you paste.
Prediction quality
Compare error, sign accuracy, a past-only historical-mean forecast, and a shuffled-symbol control. Measure uncertainty from the range of matched historical outcomes. A positive signal can still be unprofitable.
Trading reality
Execution at the observed close is not guaranteed. Slippage, spread, fees, liquidity, taxes, and overlapping positions can change returns. The trade number here is an illustration per signal, not a portfolio backtest.
Independent replication
Repeat the frozen test on later dates, several securities, and different market regimes. Publish the input data provenance, code, decisions, and all results so others can reproduce or challenge the finding.
The 78-symbol order is a real authored choice. This workbench tests one specific use of it, not all possible uses. A shuffled control probes whether this use of the order adds value. A single stock or synthetic example cannot validate a universal claim about the stock market.
Sources & research notes
Arsen Saidov, Arsik: The Limitless Handbook, “The Symbolic Layer” and “Reading the Future Through the Present” (supplied PDF).
Arsen Saidov, I Am Who I Am, “The Signal Within the Noise” and the discussion of repeating patterns and fallible prediction (English text embedded in the supplied homepage).
Arsen Saidov, 78-character sequence supplied for this research.
This page documents a research proposal. It offers no verified price forecast or investment recommendation. Future sections can add the exact algorithm, sample data, results, and revisions.
The research archive · Investigation 002
A³
Medicine & Health
Research #2 · By Arsen Saidov
Can the same discipline of encoding, comparing, and testing patterns help us ask better questions about DNA, cancer, and treatment? This chapter turns Arsen’s philosophical stone into a transparent research workflow.
Focus · Genomic comparisonStatus · Research frameworkOutcome · No treatment claims
Research #2 · Medicine & Health
01 / From the 78 symbols to four DNA bases
DNA has four principal bases, A, C, G, and T. They already appear in Arsen’s 78-character sequence. We can map each base to its position in his alphabet: A → 0, C → 2, G → 6, T → 19. This mapping is exact and reversible for those four letters. It creates a common notation; it does not itself reveal what any variant does in a living cell.
Observed DNA sequence → encode A/C/G/T → compare aligned positions → investigate differences → test biological and clinical meaning
Where Research #1 compares a chart with historical charts, Research #2 compares sequences from an appropriate reference and, in cancer research, often a tumor and matched normal sample. The analogy is useful at the comparison stage. Biology then demands more: genome position, gene and transcript context, expression, protein effect, tissue type, disease history, and independent evidence.
02 / A practical sequence laboratory
Try a short fictional pair below. The tool marks substitutions only when both sequences have the same length and are already aligned. If a comparison comes from the opposite DNA strand, select reverse complement first. It runs locally on this page and does not upload text. Use examples, not real patient sequences. The page does not save inputs, but a shared device or browser extension could still expose them.
Select “Compare sequences” to see the demonstration.
This teaching tool cannot align insertions or deletions, identify a gene, infer cancer risk, determine whether a change causes disease, or select therapy. Its position numbers begin at 1 within the pasted fragment, not at a chromosome coordinate. A clinical variant requires a genome assembly, transcript, quality-controlled sequencing, and expert interpretation.
03 / Read the output correctly
A substitution is one base changed at an aligned position. A transition swaps A↔G or C↔T; a transversion is another base swap. Those labels describe chemistry and do not grade danger. The tool reports fragment coordinates only. It cannot detect an insertion, deletion, copy-number change, gene fusion, expression change, or a tumor subclone from two pasted strings.
Why orientation matters: DNA strands are complementary. A pairs with T, and C with G. A sequence read from the reverse strand must be reversed and complemented before position-by-position comparison. The orientation control performs that transformation, but cannot determine the correct orientation by itself. NHGRI explains base pairing.
04 / From a difference to a medical question
STEP 1 · MEASURE
Generate quality-controlled data from the relevant sample and document the reference genome, tumor type, normal comparator, and laboratory method.
STEP 2 · ANNOTATE
Identify variants, gene context, and molecular effects; compare evidence in curated sources such as ClinVar. Distinguish inherited changes from tumor-acquired changes.
STEP 3 · EVALUATE
Ask whether a biomarker has a validated diagnostic, prognostic, or treatment connection for that particular cancer and patient context.
STEP 4 · EXPERIMENT
Test a biological mechanism in appropriate laboratory models. A computational match is a hypothesis, not a treatment result.
STEP 5 · CLINICAL STUDY
Evaluate safety and effectiveness in rigorously designed studies; compare outcomes, harms, and patient selection against existing care.
STEP 6 · UPDATE
Record negative and positive results, resistance, and new evidence so the interpretation improves without hiding failures.
05 / What medicine already shows
Precision oncology provides a real example of the overall approach. Biomarker testing can identify features that help clinicians choose certain cancer treatments, and some therapies have companion diagnostic tests. At the same time, not every tumor has a useful biomarker, a targeted option, or a durable response. Different cancers—and patients with the same cancer—can respond differently. Research #2 seeks to organize questions and evidence across those differences; it cannot derive a universal cure from an alphabet alone.
Finding so far
Demonstrated: Arsen’s sequence can encode A/C/G/T and show substitutions in aligned example fragments. Supported by medicine: genomic and other biomarker information can guide some cancer care when evidence and clinical context support it. Unproven: that this 78-character mapping discovers any new cure, predicts a patient’s response, or treats all cancers.
06 / The next reproducible study
Choose one cancer type and one defined research question, such as whether a specific measured variant predicts response to a specified therapy.
Use consented, de-identified datasets with clear provenance; separate discovery and untouched validation cohorts.
Define the encoding and comparison rule before looking at clinical outcomes. Compare against standard variant annotation, random or shuffled encodings, and existing clinical predictors.
Assess sensitivity, specificity, calibration, external replication, and patient-relevant outcomes, including adverse effects. Have clinicians and molecular specialists review the result.
Publish the protocol, code, uncertainty, and negative findings. A promising signal moves to biological and clinical validation; it does not become treatment advice from this webpage.
07 / A research record someone else can reproduce
INPUTS
Record sample origin, consent and access terms, tumor type, matched normal method, collection date, sequencing platform, quality thresholds, reference genome assembly and strand.
VARIANT IDENTITY
Record chromosome coordinate, reference and alternative alleles, transcript version, standardized variant representation, and uncertainty. A fragment’s “position 10” cannot be looked up clinically without this context.
EVIDENCE
Separate analytical validity, biological mechanism, clinical association, and treatment benefit. Note study type, sample size, effect size, uncertainty, disagreements, publication date, and whether findings replicate.
OUTCOMES
Prespecify a patient-relevant endpoint and comparator, then evaluate performance in an independent cohort. Record harms and resistance, not only tumor response.
CONTROLS
Compare the 78-symbol representation with ordinary A/C/G/T encoding and a shuffled character mapping. If changing the labels does not change outcomes, no unique advantage of the exact order has been shown.
BOUNDARIES
Protect genomic privacy and access rights. Separate educational code from regulated clinical interpretation and treatment decisions.
The public NCI Genomic Data Commons offers cancer research datasets, with access rules varying by data type. GA4GH variation standards address how to describe variants consistently across systems. ClinGen somatic curation evaluates the evidence for cancer-variant clinical significance. These are practical starting points for a defined study, not a substitute for reviewing the original data and clinical context.
Real example of the path: the NCI-MATCH trial assigned people with advanced cancer to study arms based on tumor molecular changes and evaluated targeted treatments. It illustrates what is required after a genomic pattern is found: specific treatments, eligibility, endpoints, safety, and observed outcomes.
Educational research prototype. Real genomic data is sensitive, and medical decisions require qualified clinical care and validated testing.
Research #3 · By Arsen Saidov
Technology
Can the philosophical stone become a usable interface for searching, building, and improving software? This section takes the same journey as the market and health chapters: represent → compare → build → test.
Representation
Use the 78-symbol alphabet as a controlled notation. Text outside it needs a declared encoding, such as UTF-8 bytes written as two hexadecimal characters each.
Retrieval
Index a defined document collection, search it, preserve citations and dates, and separate retrieved facts from generated suggestions.
Construction
Give an AI model a specification, tests, and relevant retrieved materials; run the software and compare it with an existing implementation.
Try a reversible byte encoding
This local tool converts entered text to UTF-8 bytes, then writes those bytes with 0–9 and A–F from Arsen’s alphabet. It proves an encoding property, not that the alphabet contains the meaning of the text.
Research protocol
Specify a task and baseline software. Freeze a test set before implementation.
Define inputs, output format, error behavior, privacy rules, and source provenance.
Build with the sequence as a declared encoding or ordering rule; compare with standard UTF-8 and a reordered alphabet.
Measure correctness, retrieval precision, latency, accessibility, maintainability, and failures on unseen tasks.
Scientific parallels: the Transformer paper studies learned operations on sequences; dense passage retrieval studies evidence retrieval. Neither shows that an alphabet by itself supplies all knowledge.
Research #4 · Textual and philosophical study
Soul & the Otherworld
The Torah offers language for life, breath, and human responsibility. This chapter asks how Arsen’s symbolic system can help organize a careful reading of those texts and later Jewish interpretations.
Torah text
Genesis 2:7 describes the human receiving the breath of life and becoming a living being (nefesh chayyah). Read the Hebrew and several translations before assigning a modern meaning to “soul.”
Choice in life
Deuteronomy 30:19 calls on people to choose life. It is a textual anchor for ethics and action in the present world.
Later interpretation
Jewish discussions of olam ha-ba, the world to come, draw on a wider interpretive tradition. Do not present those later descriptions as a detailed map explicitly laid out in the Five Books of Moses.
How to research the claim
Quote a short passage with its book and verse; compare Hebrew terms and translations.
Distinguish Torah, other Hebrew Bible books, rabbinic commentary, and Arsen’s own interpretation.
Ask which claim is textual, which is theological, and which could be studied empirically.
Record alternative readings respectfully. The 78-character sequence can index interpretations; it cannot experimentally verify an afterlife.
Read the primary texts: Genesis 2:7 and Deuteronomy 30:19. For a classical interpretation, see Ramban on Genesis 2:7. This section does not claim a scientific measurement of the soul or otherworld.
Research #5 · Open method
Replicate the Research
A visitor can repeat the method without Arsen present. The sequence is public; a credible result also needs a precise question, accessible evidence, frozen rules, controls, and a record of failures.
Start: copy the exact 78-character alphabet and count 78 unique symbols.
Choose: write one narrow question and define a result you could measure.
Collect: use permitted, documented data; note dates, versions, consent and missing records.
Encode: specify how raw observations map to symbols and how decoding works.
Compare: define a fixed distance or search rule and a simple baseline.
Freeze: choose parameters before an untouched validation set is revealed.
Test: run on unseen data, a shuffled-symbol control, and relevant standard methods.
Audit: report uncertainty, errors, adverse effects, costs, and alternate explanations.
Share: publish enough source detail and code to reproduce the result, with private data protected.
Revise: keep a versioned research log; new ideas become new hypotheses, never retroactive proof.
Make your own research card
This card stays in your browser. It is a starting protocol, not an automated scientific discovery.
Research #6 · By Arsen Saidov
Education
Can sequence-based feedback help someone learn?
Encode
Encode lessons and learner responses; compare progress over time while keeping human judgment in the loop.
Test
Pilot one lesson with a pre-test and later transfer test; compare an AI-assisted version with the same lesson without assistance. Measure understanding and accessibility, not just completion.
Limit
AI can draft feedback but cannot infer learning from clicks alone. Protect student data and check whether feedback helps different learners.
Starting source: UNESCO guidance. This chapter is a research proposal; future versions should record actual data, code, results and independent review.
Research #7 · By Arsen Saidov
Language & Translation
Can a shared notation preserve meaning across languages?
Encode
Represent source text, translation, context and ambiguity separately; record which translation choices are editorial.
Test
Choose a small bilingual corpus, compare human translations with AI output, and ask fluent reviewers to score fidelity, tone and culturally important omissions.
Limit
Reversible byte encoding preserves characters, not meaning; synonyms and context require human interpretation.
Starting source: Transformer translation research. This chapter is a research proposal; future versions should record actual data, code, results and independent review.
Research #8 · By Arsen Saidov
Climate & Earth
Can encoded observations help compare environmental futures?
Encode
Describe measurements, locations, units and time periods consistently; attach uncertainty and scenario assumptions.
Test
Take a public climate indicator, freeze a period, compare a simple trend with a published model’s hindcast, and report regional errors.
Limit
A symbol sequence cannot replace physical observations, emissions scenarios or climate models.
Starting source: IPCC AR6 synthesis. This chapter is a research proposal; future versions should record actual data, code, results and independent review.
Research #9 · By Arsen Saidov
Cities & Systems
Can a city be described as a testable network?
Encode
Map transit, energy, water and public space as connected data with timestamps and geographic coordinates.
Test
Prototype one neighborhood question, such as travel times to essential services, and compare modeled routes with observed journeys.
Limit
Avoid declaring an optimum without accessibility, equity, maintenance and resident input.
Starting source: UN Sustainable Development Goal 11. This chapter is a research proposal; future versions should record actual data, code, results and independent review.
Research #10 · By Arsen Saidov
The Inner Cosmos
How much of a world can a model reconstruct?
Encode
Define simulated entities, physical rules, initial conditions and an observation interface. The 78 symbols can encode the model specification.
Test
Build a small gravitational simulation, conserve a measured quantity approximately, and compare a predicted trajectory with observations.
Limit
A model reproduces chosen features at finite precision; it does not contain the entire physical universe.
Starting source: Review of cosmological simulations. This chapter is a research proposal; future versions should record actual data, code, results and independent review.
Research #11 · By Arsen Saidov
Art & Imagination
Can a finite alphabet open an unlimited creative space?
Encode
Use the sequence as a compositional score for text, sound or visual geometry; make the mapping public and deterministic.
Test
Generate works under a fixed rule and a shuffled control; ask audiences which pieces are coherent or moving without telling them which is which.
Limit
Aesthetic response is contextual and subjective; a pleasing work is not evidence of a cosmic code.
Starting source: Smithsonian Open Access. This chapter is a research proposal; future versions should record actual data, code, results and independent review.
Research #12 · By Arsen Saidov
Ethics & Stewardship
Who should benefit when a system scales?
Encode
Attach consent, provenance, attribution, access and correction paths to every dataset and tool.
Test
Audit one prototype for excluded groups, privacy exposure, false claims and how a user can contest an output. Record fixes and retest.
Limit
Technical performance alone does not justify deployment in health, finance or public life.
Starting source: NIST AI Risk Management Framework. This chapter is a research proposal; future versions should record actual data, code, results and independent review.
Research #13 · By Arsen Saidov
The Living Archive
How does research keep learning after publication?
Encode
Every new finding receives a date, version, source, hypothesis, test, result and revision. Visitors can repeat the experiments without a live global counter.
Test
Re-run one earlier claim when new data arrives and show what changed. Keep failed tests visible so the archive can correct itself.
Limit
This archive invites contributions; it does not automatically validate claims or promise autonomous discoveries after the page closes.
Starting source: Center for Open Science preregistration. This chapter is a research proposal; future versions should record actual data, code, results and independent review.
Your constellation · 0 signals
Click or press Enter on the field. Signals exist only during this page visit and are not a global visitor count.
