iPulse AIPerformance & strategiesHISTORICAL PERFORMANCE

OPEN RESEARCH. MEASURABLE OUTCOMES.

Conviction meets
the market.

What happens after the forecast?
Follow five rule-based portfolios, compare them with the market, and inspect every move.

See exactly how we measure
Market data through 30 Sept 2026

Publishing Approver: Russlan Ramdowar
Updated on

THE BIG PICTURE

What $100,000 became.

One starting balance. Five strategies. The same market.

1 Dec 2025 — 30 Sept 2026 · 7 batches · USD · Cumulative returns
Toggle any line
COMPARE WITH

Returns cover the dates above, after modeled dividend withholding and before trading and borrowing costs. SPY and QQQ are market references with different exposures; a higher return alone does not establish better risk-adjusted performance. The interest rate is a hypothetical annual assumption. Colored lines represent portfolios; thicker neutral lines and endpoint shapes identify benchmarks. The dashed $100,000 line shows holding cash with no interest.

PORTFOLIO WEALTH USD · includes cash and net dividend receivablesHover to inspect
$80k$100k$120k$140k$160kB1B2B3B4B5B6B71 Dec31 Jan1 Apr1 Jun31 Jul30 Sept
Cash BalanceStrategies may hold cash between batches. Paid dividends and sale proceeds wait until the next batch to be invested.
$0k$50k$100k1 Dec31 Jan1 Apr1 Jun31 Jul30 Sept
Cash Balance excludes reserved short collateral. Batch markers show rebalances. SPY and QQQ keep dividends in cash throughout; the interest and uninvested-cash reference lines appear only in the wealth chart.

How the research evolved

Models, synthesis and research capabilities · latest batch first
At a glanceBatch 7LatestV7 · 21 Sept 2026Batch 6V6 · 5 Jul 2026Batch 5V5 · 3 Jun 2026
Assets in the run426373312
Model familyGemini 3.8 Flash, Claude Opus 5 and GPT-6 AstraGemini 3.1 preview configurationsGemini 3.1 preview configurations
Research capability depthDocumented layers, not model IQ5 / 5 capabilities4 / 5 capabilities4 / 5 capabilities
Consensus synthesisEvidence-weightedEqual-weightEqual-weight
Explore the full comparisonAll 7 batches · research inputs, models and improvements
Research inputBatch 7LatestV7 · 21 Sept 2026Batch 6V6 · 5 Jul 2026Batch 5V5 · 3 Jun 2026Batch 4V4 · 3 May 2026Batch 3V3 · 11 Apr 2026Batch 2V2 · 19 Mar 2026Batch 1V1 · 28 Nov 2025
Assets in the run426Analyzed assets373Analyzed assets312Analyzed assets312Analyzed assets246Analyzed assets209Analyzed assets203Analyzed assets
Model familyGemini 3.8 Flash, Claude Opus 5 and GPT-6 AstraGemini 3.1 preview configurationsGemini 3.1 preview configurationsGemini 3 Pro preview configurationsGemini 3 Pro preview configurationsGemini 3 Pro previewGemini 2.5 GA and Gemini 3 preview configurations
Opinions per assetOriginal 1-year ranking inputs14Includes advisor modes and configurations12Intended advisor modes and configurations
Details12 opinions intended per asset; a few assets have 11 completed opinions because one output is missing.
12Intended advisor modes and configurations
Details12 opinions intended per asset; a few assets have 11 completed opinions because one output is missing.
10–12Includes advisor modes and configurations10–12Includes advisor modes and configurations4–16Includes advisor modes and configurations10–12Includes advisor modes and configurations
Research capability depthDocumented layers, not model IQ5 / 5 capabilities4 / 5 capabilities4 / 5 capabilities2 / 5 capabilities2 / 5 capabilities1 / 5 capabilities1 / 5 capabilities
Portfolio statusTrades availableSimulated activity · inspect the ledger belowTrades availableSimulated activity · inspect the ledger belowTrades availableSimulated activity · inspect the ledger belowTrades availableSimulated activity · inspect the ledger belowTrades availableSimulated activity · inspect the ledger belowTrades availableSimulated activity · inspect the ledger belowTrades availableSimulated activity · inspect the ledger below
Global contextEvents between the model’s knowledge cutoff and the forecast run dateYes
DetailsCombined context coverage: 1 Jan 2025 – 20 Sept 2026.Governed world-state context accompanies the individual forecasts.
Yes
DetailsCombined context coverage: 1 Jan 2025 – 31 May 2026.Governed global context was supplied.
Yes
DetailsCombined context coverage: 1 Jan 2025 – 10 Apr 2026.Governed world-state context was included.
Yes
DetailsCombined context coverage: 1 Jan 2025 – 10 Apr 2026.Governed global context was included in the run.
Yes
DetailsCombined context coverage: 1 Jan 2025 – 10 Apr 2026.Verified 2025–2026 world-state context was supplied. This is the first batch for which that claim is valid.
No
DetailsBatch 2 did not receive the later world-state context files.
No
DetailsNo global-context component was supplied to Batch 1.
Asset fundamentalsEarnings reports, balance sheets and key financial ratiosEquities only
DetailsCurrency-normalized fundamentals for eligible equities.
Equities only
DetailsCurrency-normalized fundamentals were supplied for eligible equities.
Equities only
DetailsAsset-specific fundamentals were introduced for equities, not for asset classes without company financial statements.
No
DetailsAsset-specific fundamentals were not yet part of the run.
No
DetailsAsset-specific fundamentals were not yet part of the run.
No
DetailsNo governed fundamentals input was supplied.
No
DetailsNo governed fundamentals input was supplied.
Cross-asset rankingsYes
DetailsPortfolio simulations use the frozen original quantitative leaderboard, separate from the synthesized forecast path.
Yes
DetailsCurrent v7.4 architecture adds event-risk pressure and equity financial health; historical snapshots retain their formula identifiers.
Yes
DetailsThe v7.2/v7.3 ranking generation incorporated dividend, dispersion, volatility, and cash-hurdle logic.
Yes
DetailsRankings used the then-current post-generation formula and were later eligible for reproducible re-scoring.
Yes
DetailsPost-generation ranking work existed, with formulas continuing to evolve after the run.
Added later
DetailsThe first ranking implementation was created after this run, beginning March 27, 2026.
Added later
DetailsRanking did not exist at generation time; frozen forecasts were eligible for later retrospective scoring.
Consensus synthesisEvidence-weightedIntelligence layer · explained, adjustable opinion weights
DetailsThe synthesizer assigns explained per-opinion weights; Python calculates and compounds the weighted quarterly returns.
Equal-weightSimple averaging · every opinion counts equally
DetailsIndependent forecasts were synthesized into asset-level consensus records.
Equal-weightSimple averaging · every opinion counts equally
DetailsA second batch layer combined independent advisor forecasts into an asset consensus.
No
DetailsThe dedicated asset-consensus layer followed this batch.
No
DetailsThe dedicated asset-consensus layer had not yet been introduced.
No
DetailsThe asset-consensus layer had not yet been introduced.
No
DetailsForecasts were produced independently without the later asset-consensus layer.
Editorial reviewYes
DetailsSeparate editorial publication approval remains downstream of forecasts and rankings.
Added later
DetailsThe formal editorial-review workflow was added after Batch 6 generation and now governs publication of Editorial Top Picks.
No
DetailsThe formal editorial stage was introduced after the next deep-analysis execution.
No
DetailsThe editorial publication workflow did not yet exist.
No
DetailsThe editorial publication workflow did not yet exist.
No
DetailsThe editorial publication workflow did not yet exist.
No
DetailsThe editorial publication workflow did not yet exist.
Model diversity comparisonYes
DetailsMatched Universal Investor researcher configuration across Flash, Opus and Astra; each asset has 12 Flash, one Opus and one Astra opinion.
No
DetailsThis batch predates the matched three-model comparison.
No
DetailsThis batch predates the matched three-model comparison.
No
DetailsThis batch predates the matched three-model comparison.
No
DetailsThis batch predates the matched three-model comparison.
No
DetailsThis batch predates the matched three-model comparison.
No
DetailsThis batch predates the matched three-model comparison.
Evidence-based opinion weightingYes
DetailsBlinded opinion identities support lower or zero weights for weak claims and higher weights for supported insights. Weights are normalized and the complete source path is frozen.
No
DetailsNo Batch 7 per-opinion judgment allocations were applied.
No
DetailsNo Batch 7 per-opinion judgment allocations were applied.
No
DetailsNo Batch 7 per-opinion judgment allocations were applied.
No
DetailsNo Batch 7 per-opinion judgment allocations were applied.
No
DetailsNo Batch 7 per-opinion judgment allocations were applied.
No
DetailsNo Batch 7 per-opinion judgment allocations were applied.
What changed
3 improvements
  • Compare three model families using the same researcher framework.
  • Replace equal opinion contribution in the synthesized path with explained, normalized judgment weights.
  • Preserve both averaged and synthesized consensus; weighting remains experimental and does not establish better accuracy.
2 improvements
  • Aligned reporting and ticker currencies and corrected P/E and related fundamental calculations.
  • Expanded the analyzed asset universe and shortened selected narrative fields.
4 improvements
  • Added asset-level consensus synthesis after independent advisor forecasts.
  • Added asset-specific fundamentals after global context for equity assets.
  • Standardized report language at subject-matter-expert level and added glossary assistance.
  • Increased Thinker-mode use and reduced overly long output fields.
2 improvements
  • Repeated the task goal and critical asset instructions at both ends of long prompt assemblies.
  • Kept global context while reducing the risk that essential instructions were lost inside 50,000-plus-token prompts.
3 improvements
  • Added global events and quantitative world-state context, arranged to preserve temporal proximity.
  • Expanded the asset universe and further revised the structured output schema.
  • Removed the ambiguous Nostradamus persona configuration.
3 improvements
  • Added persona, investment-framework, behavior, and communication-style configuration.
  • Tightened forecast-anchor and horizon-step requirements in output schema v4.
  • Standardized the main forecast path around a five-year structure.
2 improvements
  • Introduced machine-readable structured JSON outputs for comparison and later inspection.
  • Ran separate three-year and five-year forecast configurations as an early horizon experiment.

How the five-brain scale works: one point each for independent forecasts, global context, equity fundamentals, consensus synthesis and evidence-based opinion weighting. Partial equity coverage counts; features added after a run do not. This is a capability checklist—not measured intelligence, accuracy or investment performance.

Batch 7 adds matched Flash, Opus and Astra comparisons and explained synthesizer weights. Earlier synthesis used equal opinion weights. These are research methods, not evidence of improved accuracy. Portfolio simulations still use frozen quantitative rankings, separate from the synthesized forecast path. Engine history Lessons learned

A reconstruction, not a live track record. These portfolios apply frozen batch rankings retrospectively from forecast dates. Some scores were archived later. Returns are hypothetical, net of modeled dividend withholding, before trading and borrowing costs. Read the methodology

BEYOND THE LINE

Every position has a story.

Explore one portfolio’s holdings and complete transaction record.

10 buys · weighted

Buy the 10 highest-ranked eligible stocks. Give higher ranks more capital. Sell a position after its price gains 25%; hold the proceeds until the next batch.

Allocation & exit rules
Portfolio wealth$124,334+24.33% since Batch 1
Cash Balance$00.0% of portfolio wealth
Largest peak-to-trough fall-13.95%Maximum drawdown in this period
Net dividends received$1,278Paid cash, after modeled withholding
Wallet -$0 − reserved short collateral -$0 = free cash $0.Unpaid dividends: $0.
Showing the latest 150 of 161 events. Load the remaining history to search, sort or export the complete record.
22 Sept 2026Batch 7NKE.NYSENKENIKE, Inc.Buy106.386-$3,841-$0
22 Sept 2026Batch 7TCS.NSETCSTata Consultancy Services LimitedBuy232.536-$5,121$3,841
22 Sept 2026Batch 7CLBT.NASDAQCLBTCellebrite DI LtdBuy548.961-$6,401$8,961
22 Sept 2026Batch 7REL.LSERELRelx PLCBuy272.959-$8,961$15,362
22 Sept 2026Batch 7HDFCBANK.NSEHDFCBANKHDFC Bank LimitedBuy1,325.45-$10,241$24,323
22 Sept 2026Batch 7ACN.NYSEACNAccenture plcBuy69.681-$12,802$34,565
22 Sept 2026Batch 7ADBE.NASDAQADBEAdobe Inc.Buy64.479-$15,362$47,367
22 Sept 2026Batch 7BSX.NYSEBSXBoston Scientific CorporationBuy398.987-$17,922$62,729
22 Sept 2026Batch 7ADYEN.AMSADYENAdyen NVBuy20.957-$21,763$80,651
22 Sept 2026Batch 7INTU.NASDAQINTUIntuit IncBuy87.578-$25,604$102,414
1–10 of 150 records · 150 of 161 events loaded · Date / batch ↓
1 / 15

TRUST COMES FROM THE DETAILS

No black box.
Here are the rules.

A rising line is only part of the evidence. Timing, risk, taxes, cash and execution assumptions all shape the result. We make those choices visible.

01 / SELECT

Start with the frozen ranking.

Use the original batch’s one-year stock ranking with at least 80% directional agreement. Each starting batch opens a fresh $100,000 model portfolio.

02 / FOLLOW THE RULES

Rebalance. Exit. Wait.

At every new batch, reallocate the portfolio using the strategy’s rules. Dividends and exit proceeds stay in cash until then. The allocation cap is 35% per underlying asset.

03 / ACCOUNT FOR EVERYTHING

Track wealth and spendable cash.

Value positions in USD. Include net paid dividends and unpaid entitlements; deduct short liabilities. Separate short-sale collateral from money available to invest.

Five strategies, explained.

The same rules apply across all selected batches.
01Top 10 · weighted, +25% exitSell at +25%

Buy the 10 highest-ranked eligible stocks. Give higher ranks more capital. Sell a position after its price gains 25%; hold the proceeds until the next batch.

Allocation: 20%, 17%, 14%, 12%, 10%, 8%, 7%, 5%, 4%, 3% by rank. If the eligible list is shorter, weights are redistributed within the 35% cap; any unallocatable balance stays in cash. Concentration trims may occur between batches.

02Top 10 · equal, +25% exitSell at +25%

Buy the 10 highest-ranked eligible stocks with equal starting allocations. Take profits after a 25% price gain and wait for the next batch to reinvest.

Allocation: 10% per stock when all 10 are eligible. If the eligible list is shorter, weights are redistributed within the 35% cap; any unallocatable balance stays in cash. Concentration trims may occur between batches.

03Top 10 · weighted, holdRebalance each batch

Buy the 10 highest-ranked eligible stocks, with more capital in higher ranks. Keep positions between batches, then rebalance to the new Top 10.

Allocation: 20%, 17%, 14%, 12%, 10%, 8%, 7%, 5%, 4%, 3% by rank. If the eligible list is shorter, weights are redistributed within the 35% cap; any unallocatable balance stays in cash. Concentration trims may occur between batches.

04Top 5 · forecast-target exit130% of forecast upside

Buy the five highest-ranked eligible stocks. Exit when the gain reaches 130% of the original one-year forecast upside. At the next batch, retain and resize overlapping names and refresh their targets.

Allocation: 20% per stock. A 20% forecast gain gives a 26% exit target, not a 130% price gain. If the eligible list is shorter, weights are redistributed within the 35% cap; any unallocatable balance stays in cash. Concentration trims may occur between batches.

05Top 20 buys + 5 shortsRebalance each batch

Allocate 60% to the 20 highest-ranked eligible buys and 40% to short positions in the five lowest-ranked eligible sells. Rebalance at each batch. Shorts profit when prices fall and lose when prices rise.

Allocation: 3% per buy and 8% per short when all names are eligible. Short proceeds are reserved as collateral; they are not free cash. If the eligible list is shorter, weights are redistributed within the 35% cap; any unallocatable balance stays in cash. Concentration trims may occur between batches.

Inspect the evidence. Form your own view.iPulse AI is an open agentic investment research platform. These experiments support research; they do not establish future returns or replace your investment judgment.

Explore our methodology

What these results do—and do not—tell you.

Historical reconstruction and timing

This view starts on the first execution date after each original forecast date. Some original scores were archived after those dates, so this reconstruction does not establish that these exact portfolios were available to trade at the time. It is retrospective research, not an audited live account or out-of-sample proof. The featured set was chosen after historical results were available; selection bias is possible.

Prices, currencies and dividends

The calculations use saved daily prices, FX and corporate actions. Stock and dividend currencies are handled separately, including GBX-to-GBP conversion. Dividends create entitlements on the ex-date and spendable cash on the payment date, after modeled issuer withholding. Unknown payment dates remain unpaid. Net receivables count toward wealth but cannot fund purchases. Source corrections and alternate-venue quote assumptions are retained in the underlying evaluation.

Benchmarks and comparison basis

The S&P 500 comparison buys SPY and the Nasdaq-100 comparison buys QQQ once at the selected starting batch and holds them; net dividends accumulate in cash. Both are ETF proxies, not official total-return indices. The cash line compounds the editable annual rate over elapsed calendar days using a 365.25-day year; it is a hypothetical constant-rate benchmark, before personal tax. Strategy cash does not earn this benchmark rate.

Execution, costs and concentration risk

Prices and exits use saved daily observations, not a guaranteed intraday limit fill. No commission, slippage, capital-gains tax or short-borrowing fee is modeled. Shorts assume availability without broker margin constraints and owe dividends. The 35% cap is checked at daily valuations and executable closes; prices can move between observations. Returns are period returns, not annualized forecasts or a promise of future results.

Source, freshness and reproducibility

This view uses a verified evaluation generation with prices through 30 Sept 2026, frozen original leaderboard batches, and independently initialized reports for every starting batch. Each generation uses saved prices and original leaderboard data that passed the evaluation checks. Each view stays on one verified generation while you explore; the as-of date identifies the prices used. Table exports include all records matching the current strategy, start batch and filters.