★ Synthos One

name: Synthos One

as_of: 2026-07-04

holdings: 100

buckets: {core: 50.0, tactical: 30.0, strategic: 15.0, crypto: 5.0}

schema: synthos-one-v1

author: Fable, Chief Investment Architect

sources_of_truth:

- docs/fep_active_inference.md

- docs/fep_buildout_fable.md

- results/macro/regime_2026-07-04.md

- results/flagship/2026-07-03.md

- results/crypto/flagship_portfolio.md

- results/crypto/scored.json

- docs/crypto_framework_fable.md

- results/deep_dives/*/2026-07-03.md (front matter: verdict, R/G/E scores, fair value)

100 positions · stocks + ETFs + crypto · Free-Energy-Principle-driven · core-satellite across three clocks

Synthos Research · 2026-07-04 · Hypothetical model portfolio. NOT a track record. NOT investment advice. Honesty is the product.

0. THE HONEST PREMISE — READ THIS FIRST

What we will not promise. We will not promise 30% a year, and we will not promise 10x in five years. Sustained 30% CAGR is Peter-Lynch-at-Magellan territory — roughly the best sustained public-market record ever produced, achieved by one person, once, over 13 years. 10x in 5 years requires ~58% compounded annually — for a diversified 100-position portfolio that is essentially unprecedented in recorded market history. Anyone selling you that number from a broad book is selling you a story.

What the target actually is. The exponential ambition is a north star that informs construction — it is why the satellites hunt forward "next exponentials" instead of trailing compounders, why the crypto tail exists, and why sectors the regime punishes get zero dedicated capital. It is a design principle, not a forecast.

What the proof actually is. The only honest proof is the calibrated track record: every regime posterior, tier rating, and position is logged pre-publication, graded at maturity by the same non-gameable grader that scores the pundits (data/thinker_scores.json — 422 matured claims graded to date, 296 correct / 126 incorrect, 70.1% overall; the engine's own calls sit in the same file under synthos_self: 137 positions seen, 0 matured, status pending, explicitly not back-fitted). Until that record matures, Synthos One's performance claims are exactly zero. Section 9 states the realistic return range and what would have to go right.

The breadth-vs-return tension, resolved in the open. A 100-ticker portfolio cannot 10x — breadth is beta, and beta is ~10% a year, not 58%. A 5-ticker portfolio could 10x — and could also destroy you, and its result would be luck as much as skill. Synthos One resolves the tension structurally, not rhetorically, with core-satellite:

Half the book guarantees you roughly the market. The other half is where every unit of research edge is spent. If the edge is real, the satellites add points over beta; if it isn't, the calibration log will say so in public.

1. THE REGIME CALL THAT DRIVES THE TILTS

From results/macro/regime_2026-07-04.md (v2, enriched — the belief layer's current top-level output):

Reflation tipping toward Inflation, under structural Fiscal Dominance, spined by a demographic debt supercycle.

The load-bearing numbers, all cited from that report:

• Growth firm: industrial production +1.7% YoY and accelerating; retail sales +6.9% YoY; new orders +10.4% YoY (the AI/reindustrialization capex wave in hard data). No recession in the data.
• Inflation re-accelerating: headline CPI 4.3% YoY, core PCE 3.4% — six-month impulse positive on every price series. Forward Guidance: core floors near 3%, not 2%, on 5–6% deficits.
• Fiscal dominance: debt/GDP 122.6%, deficit -5.77% of GDP at full employment, interest outlays +41.5% YoY ($1,219B). Alden's mechanism: above ~100% debt/GDP, rate hikes inflate the deficit — the tool is broken.
• Demographics lock it in: fertility 1.63, old-age dependency 28.5 (vs 14.7 in 1960), LFPR falling (-0.9 YoY) into job openings +3.9% — labor supply is the binding, inflationary constraint for a decade.
• The forward vector — the liquidity turn: RRP effectively empty ($2.17B), bank reserves -4.8% YoY, M2 already +5.6% — the mechanical setup that historically forces QT to end. Visser sizes the eventual injection at $6–8T. Direction: liquidity turns UP within 6–18 months; timing unforecastable.
• Bond market: 30y at 4.97% testing Gromen's ~4.8% affordability ceiling; base case bear-steepener; HY spreads 275bp = zero credit stress today.

Positioning consequence (regime report §5, verbatim logic): own real assets and pricing power; rent the front end; don't lend long to a borrower who is demographically compelled to debase. That sentence writes the Tactical sleeve and every 0% in Section 4.

2. CONSTRUCTION — HOW THE 100 WERE CHOSEN

• CORE (50%, 19 positions). Broad-market ETFs (SPY, QQQ, RSP) as the representative base, plus every Buy — Core name from the equity flagship (results/flagship/2026-07-03.md — the 13 Buy—Core verdicts across the S&P 500/Nasdaq-100 deep-dive sweep) and three Buy—Tactical mega-quality names the flagship top-weights (GOOGL, AMZN, LLY). RSP (equal-weight S&P) is deliberate: Mag7 is guilty-until-proven, so the index base itself de-concentrates.
• TACTICAL (30%, 45 positions). The regime report's tilts, expressed in ETFs (GLD/GDX/SLV/PDBC/XLE/BIL) plus the Buy—Tactical equity roster where it aligns with the regime: energy, miners/materials, steepener-and-fee financials, pricing-power healthcare. Sectors the regime punishes get zero dedicated capital, each justified in writing (Section 4).
• STRATEGIC (15%, 16 positions). The secular next-exponentials: AI-capex supercycle (Pal: "the largest capex supercycle in human history"; compute-vs-energy) and its power/space adjacencies. Names sourced from the deep-dive scores: every holding has Exponential Potential ≥ 7/10 (or Buy verdict with E≥6). Honesty note: several carry Hold verdicts at spot price — the verdicts are price-disciplined, so this sleeve is tranche-entry only (Section 7); the Hold verdict is precisely why these are satellites, not core.
• CRYPTO (5%, 20 positions). The crypto flagship (results/crypto/flagship_portfolio.md, built from the 73-token scored.json) scaled by exactly 0.05 — same names, same proportions, BTC-anchored at 30% of the sleeve. A 100% loss of the sleeve costs 5.0% of Synthos One.

No ticker appears twice. GOOG (the flagship's second Alphabet line) is consolidated into GOOGL.

3. THE FULL 100 — TICKER TABLE

Weights sum to 100.00%. Tier/score citations: equity R/G/E = Downside Risk / Growth Quality / Exponential Potential (0–10, from deep-dive front matter, 2026-07-03); crypto = composite/tier from scored.json (2026-07-04).

Top ~15 by weight (everything ≥ 2.00%)

CORE — 50.00% (19 positions, multi-year clock)

TACTICAL — 30.00% (45 positions, 6–18mo clock — the regime sleeve)

Every line here is a consequence of Section 1. The sleeve is deliberately wide-and-shallow: the bet is the REGIME, expressed through many correlated-but-distinct lines, so no single name can invalidate the macro call's P&L.

Real assets / hard money / front end (6 ETFs, 8.75%):

Energy equities (7, 5.00%) — all Hold verdicts at spot; owned as a sector regime tilt with entry discipline, sized small per line:

Miners & materials (4, 3.25%):

Financials — steepener & risk-on-credit beneficiaries (9, 4.50%) — all Buy—Tactical verdicts from the sweep:

Healthcare with pricing power (5, 3.50%) — all Buy—Tactical:

Regime-consistent industrials, tech & consumer (11, 4.25%) — Buy—Tactical roster, small lines:

Payments/fintech value (3, 0.75%) — Buy—Tactical trio, deliberately tiny:

STRATEGIC — 15.00% (16 positions, 3–5yr clock — the next-exponentials sleeve)

Selection rule: Exponential Potential ≥ 7 in the 2026-07-03 deep-dive sweep (or Buy verdict with E≥6). Pal's frame: AI is "the largest capex supercycle in human history; bottlenecks in chips and power only force MORE expenditure, not less." This sleeve owns the bottlenecks. Honesty: most carry Hold verdicts at spot — that is the price discipline talking, so this sleeve enters in tranches only (Section 7) and is sized as satellite, never core.

CRYPTO / CONVEXITY — 5.00% (20 positions, BTC-anchored)

The crypto flagship (results/crypto/flagship_portfolio.md) scaled by exactly 0.05. Full per-token theses, entry playbooks, and unlock-calendar rules live in that document — it is the governing sub-portfolio. Composite/tier from scored.json (73-token universe). Crypto is the purest liquidity-cycle expression (framework §4.2), and the liquidity turn is the regime report's highest-conviction forward vector — this sleeve is that sentence, sized for total-loss survivability.

4. SECTOR ALLOCATION — WITH THE ZEROS JUSTIFIED IN WRITING

Dedicated (satellite + named-stock) capital by sector. The broad index base (SPY/QQQ/RSP = 16.0%) carries every sector at roughly market weight underneath this table — so a 0% row means zero dedicated capital and zero conviction, not a net-short. Synthos One is never net-short a sector.

Roll-up: equities ~62% named + 16% index base, real assets ~8.75% (GLD/SLV/GDX/PDBC + NEM inside materials), crypto 5%, cash 1.75%.

5. THE FEP PIPELINE — HOW BELIEFS BECOME WEIGHTS (OPERATIONAL)

This is the build-out I specified in docs/fep_buildout_fable.md, now wired to portfolio construction. The one-sentence version (framework doc): maintain an explicit probabilistic world model, update it by precision-weighted Bayesian inference from the KB + data, emit calibrated distributions, and act on expected return traded off against uncertainty.

The flow

`

KB CLAIMS (~27,800 dated, per-thinker) FRED VECTOR z_t (growth, inflation,

precision κ_c = w_job × m_score × recency liquidity, credit accelerations)

(thinkers.py job routing × scorecard κ_z = 1.0 — data is the anchor;

shrinkage × recency.py decay; claim mass capped at Σκ_c ≤ 1.0

e.g. Dale/Steno tactical 1.9, Pal so the commentariat can match the

strategic 1.9, Visser selection 2.0, data's vote but never outvote it

mgmt voices 0.5)

└──────────────┬───────────────────────────┘

▼

REGIME POSTERIOR q(s), s ∈ {Goldilocks, Reflation, Inflation, Deflation}

log q(s) = log p̃(s) + κ_z·log N(z;μ_s,Σ) + Σ κ_c·log ℓ_c(s) − log Z

(buildout §3.5; prior drifts toward base rates, λ=0.02/wk — beliefs expect to go stale)

▼

SECTOR / NAME SIGNALS

• Sector tilts = posterior-weighted blend of per-regime playbooks (NOT argmax —

today's call is Reflation→Inflation, so the book blends both playbooks)

• Names = deep-dive tiers (Buy—Core/Tactical, R/G/E) + crypto composites (V/A/T/M)

• Fragility F_t = KL(q_synthos ‖ q_market)·(1−H/log4) sizes any contrarian sleeve

▼

WEIGHTS

• CORE: fixed representativeness mandate (only names change, on tier migration)

• TACTICAL: regime posterior sets sector over/underweights and the zeros

• STRATEGIC: selection-job conviction (E-scores), regime-gated on gross only

• CRYPTO: composite ranks tokens; regime sizes the sleeve ("the composite ranks

tokens; the regime sizes crypto" — crypto framework §4.2)

▼

CALIBRATION LOG (the honesty proof)

Every posterior and tier logged append-only pre-publication (buildout §3.7),

graded at maturity by the same grader as the pundits. synthos_self: 137 positions

seen, 0 matured, pending, not back-fitted. No grades, no performance claims.

`

Cadence: monthly rebalance, weekly triggers

Monthly (scheduled): recompute the regime posterior; rebalance sleeve weights to targets; apply tier migrations; crypto rebalance per its flagship rules; log everything append-only. Quarterly: full reconstitution (universe review, taxonomy review, this document reissued).

Weekly trigger list — any ONE fires an out-of-cycle review; thresholds are pre-committed so hindsight can't edit them:

What's live now vs. what needs the full environment — stated honestly

Live today (this portfolio was built from it): the distilled artifacts — the enriched regime report (posterior expressed narratively as Reflation→Inflation), 200+ equity deep-dive tiers with R/G/E scores, the 73-token crypto scored.json, the thinker scorecard (thinker_scores.json: 422 graded, 70.1%), and the job-routing weights in thinkers.py. Everything cited in this document reconciles to those files.

Needs the full KB environment (lancedb + conda biotech): live belief-layer generation — beliefs.py weekly posterior updates from fresh FRED pulls + newly distilled claims, the market-implied posterior q_market, the fragility score F_t, EIG research routing, and the append-only regime_log.jsonl calibration log. Until that runs weekly, the regime posterior is refreshed manually per report cycle, triggers 3–5 are approximated by their observable proxies (regime-report reissues, Kritzman turbulence from macro.py), and the calibration page cannot exist. We say this plainly because claiming a live belief engine before it runs weekly would be exactly the kind of dishonesty this brand exists to kill. The Phase-2 falsification rule stands: if the FEP layer doesn't beat the dumb baselines after 26 logged weeks, we rip it out and rename the docs (buildout §4–5).

6. THE PLAYBOOK — HOW TO THINK ABOUT THIS PORTFOLIO

The three-clock philosophy

The book runs on three clocks because forecast skill is horizon-specific — that is the entire lesson of thinkers.py (Steno/Dale weighted 1.9 on tactical, 0.8–1.0 elsewhere; Pal 1.9 strategic but 0.6 tactical — "don't use him for a 3–6mo call"; Visser 2.0 on selection). The portfolio routes capital the same way it routes voices:

• The multi-year clock (CORE) answers: is capitalism still compounding? You don't trade this clock; you feed it. Its enemy is your boredom.
• The 6–18mo clock (TACTICAL) answers: what does the current regime pay for? It is wrong the most often and repriced the most often — which is why it's 45 shallow lines, monthly rebalanced, and gated by pre-committed triggers instead of feelings.
• The 3–5yr clock (STRATEGIC) answers: what becomes 5x bigger regardless of next quarter? Its enemy is entry price — hence tranches — and narrative decay — hence the E-score refresh each sweep.
• The crypto tail is not a clock; it is convexity. It exists because the regime's single highest-conviction forward vector (the liquidity turn) has its purest, highest-beta expression there, and it is sized so that being completely wrong costs 5 points, not the portfolio.

How to invest in it

1. DCA the core. Equal installments over 8–12 weeks into the 19 core lines (or start with SPY/QQQ/RSP and phase the 16 stocks in). After that, automatic monthly additions. Never time this sleeve.
2. Tranche the satellites. Tactical: half position now, half on sector pullbacks. Strategic: thirds over 3–6 months, accelerating into 15–25% drawdowns of names whose E-scores hold. The deep-dive entry zones govern (e.g., NVDA $191–195, exit discipline at a close below the 200-day ~$191).
3. Crypto by its own book. The crypto flagship's per-position styles apply verbatim: DCA BTC/ETH/LINK; tranche the quality DeFi; drawdown-only buys for HYPE; mechanical momentum stops for SPX/BONK — no averaging down in memes, ever.
4. Rebalance monthly to targets; reconstitute quarterly. Drift bands: ±20% relative per line, sector caps at target +5 points. Trim winners into strength; the log records every change.
5. The regime gate on gross exposure (generalizes the crypto flagship's gate to the whole book):

1. What would change the allocation (the honest pre-commitments): a deflationary break (triggers 1–2) rewrites the Tactical sleeve and re-opens Treasuries; core PCE credibly breaking toward 2% with growth intact (Goldilocks) would trim GLD/energy toward growth; the AI-capex digestion Visser flags (~$4T IPO wave, 3–6mo pause) would slow strategic tranches but not the thesis; a fragility flag with high confidence licenses the only contrarian positioning we ever do; and any tier migration executes mechanically regardless of narrative.

Rules that protect you from us

• No position except broad ETFs may exceed 8%. No single crypto line except BTC above 0.5%.
• Every 0% sector carries a written re-open condition (Section 4) — zeros are theses, not omissions.
• If the calibration record (once live) shows the regime engine underperforming the persistence baseline, the tactical sleeve's active share shrinks automatically toward sector-neutral until it re-earns it. The book de-risks its own beliefs.

7. INVESTABLE MONDAY — ENTRY APPROACH PER BUCKET

Approach and logic, not per-line price targets. A reader with the list above can act at Monday's open.

CORE (50 units of capital): Buy one-third at Monday's open across all 19 lines — representativeness doesn't wait, and the multi-year clock makes the entry price a rounding error. Schedule the remaining two-thirds as 8 weekly automatic buys. If the market gaps down >2% intraweek, pull the next installment forward. Logic: eliminate timing regret in both directions; the core's job is to exist, not to be clever.

TACTICAL (30 units): Buy half at the open for the ETF block (GLD/GDX/SLV/PDBC/XLE/BIL) and the Buy-rated equities; hold the second half for pullbacks toward each name's nearest support band per its deep-dive entry zone (the flagship convention: nearest moving-average support up to current price). For Hold-verdict energy/materials lines, first tranche at the open is a half of the half — the regime justifies presence, the verdict counsels patience. Do not chase anything that gapped >3% that morning; the regime is a 6–18 month thesis and will offer entries.

STRATEGIC (15 units): One-third at the open, one-third on a 15% pullback, one-third on 25% — per name, standing orders. These are Hold-at-spot names with E≥7: the thesis is bought immediately at one-third size so the supercycle can't run away from you, and the price discipline that produced the Hold verdicts is honored with the remaining two-thirds. Exception: the Buy-verdict lines (NVDA, ANET, DELL, HPE) start at half. Every tranche cancels if the name's next sweep drops its E-score below 6 or verdict to Sell.

CRYPTO (5 units): Execute the crypto flagship's per-position playbook starting Monday: begin the BTC weekly DCA and ETH/LINK monthly DCA immediately; place tranche-1 orders for SOL/KAS/AAVE/SKY/PYTH/POL/CAKE/JUP per its §4 table; do not chase AERO/ETHFI/GRASS/SPX (extended — pullback rules only) and do not touch HYPE except on a −25% drawdown. All unlock-calendar rules apply from day one.

8. HONEST RISKS — WHAT KILLS THIS PORTFOLIO

1. The regime call is the concentrated bet. Core is diversified; the satellites are one thesis worn forty ways. If the liquidity turn doesn't come — or comes after a deflationary accident — the tactical sleeve (energy, miners, banks) and the entire crypto sleeve fall together. The crypto flagship says it verbatim: "in a liquidity drain, every sector is one trade." Mitigations: the 50% core, the trigger table, tranche discipline. Mitigations, not cures.
2. The specific poison scenario: a deflationary bust (credit event, claims spike) in which energy, gold miners, banks, AI capex, and crypto all fall while the one asset we hold at zero — long Treasuries — rallies. Our zeros are our biggest active risk, and we accept them knowingly because the regime evidence (fiscal dominance, demographics, 4.3% CPI) says duration is the worse standing bet. The trigger table exists to buy duration late but alive.
3. Early ≠ right. "Position ahead of the injection" can mean quarters of underperformance before vindication — and in crypto, −50% before right. The 2026 tape has already cut BTC in half from ATH; it can halve again.
4. Gromen's tail: capital-flight price action — dollar, bonds, and stocks down together. GLD/BTC are the intended hedges; in the first weeks of such a move, correlations go to one and nothing hedges.
5. Valuation risk in the strategic sleeve. Twelve of sixteen are Hold-at-spot. Paying up for E-scores is how forward investors die; the tranche rules and R-score sizing (CRWV R9 = 0.5%) are the containment.
6. Model risk, stated in our own words: the FEP layer could be "plain Bayesian forecasting in a Friston costume" (framework doc's own guardrail), the regime state-space is a 4-box cartoon of macro, and per-thinker precision is regime-conditional (Gromen's 0.500 on n=84 was earned in one dollar regime). The falsification test with a rip-out date is pre-committed (buildout §4.2).
7. Drawdown expectation, in numbers: a normal bad year for this construction is −20% to −35% peak-to-trough (equity beta ~1 with cyclical tilts, plus a crypto sleeve that can lose 80% of itself = up to −4 points on its own). If you cannot hold through that without selling the bottom, hold more BIL and less of everything else. Measured in gold — Gromen's own yardstick (S&P −35–40% vs gold since 4Q21) — even good nominal years may look pedestrian. We will report both numerators.

9. WHAT RETURN IS REALISTIC — AND WHAT WOULD HAVE TO GO RIGHT

The honest expected range (nominal, full-cycle annualized):

Say the quiet part out loud: 30% sustained is Lynch-tier — the single best sustained public record ever, and he ran it into a far less efficient market. 10x in 5 years is 58% compounded — no diversified 100-position book has a credible claim to it. Synthos One is built so that if the exponential tail arrives, the book participates meaningfully (15% strategic + 5% crypto + concentrated tactical tilts), and if it doesn't, the 50% core keeps the result civilized. The calibrated track record — logged pre-publication, graded by the same scorer as the pundits, published including the misses — is the only performance claim we will ever make. Today that record's honest content is: pending, n=0 matured.

#1 risk, named once more: the macro regime call itself. Every satellite leans on Reflation→Inflation + the forward liquidity turn. If that posterior is wrong, no amount of stock-picking inside the sleeves saves the year — only the core, the triggers, and the tranches do.

Synthos Research · Hypothetical model portfolio constructed 2026-07-04 from the cited desk artifacts. No real money is deployed in this construct; there is no track record; model-portfolio arithmetic is survivorship-inflated hindsight until the calibration log matures. Nothing here is personalized investment advice. Digital assets can go to zero. Do your own research.