I Used AI to Grade a Real 16-Fund Portfolio. The Result Surprised Me

What the Video Analyzes

• A real “16-fund” ETF portfolio used for retirement (held in Traditional IRA and Roth IRA, though the split is not provided).
• A comparison against a simplified three-fund portfolio created by “folding” similar exposures.
• A model of the impact of a 1.6% annual effective fee, representing an advisor charging:
  • a $10,000/year flat fee
  • plus roughly 0.69 bps (AUM-based)
  • plus underlying ETF expenses This is used to show how fees affect returns and risk metrics.

Explicit Finance Context / Cautions

• “Widow’s penalty” is highlighted as an important retirement tax-planning issue: if the surviving spouse becomes a single filer, tax brackets tighten and taxation can become less favorable. A quoted concern is that the advisor reportedly said they’d never heard of it—flagged as a potential red flag.

• The speaker repeatedly includes disclosures that they are not a financial advisor and not a registered investment advisor—framed as commentary (“two cents”).

• Fee caution: the presenter emphasizes not telling viewers to immediately fire an advisor. Instead, they argue for evaluating whether goals can be met and considering relationship/work preferences.

Instruments / Tickers Mentioned

US / Total Market

• VOO (S&P 500 ETF mentioned)
• VTI (modeled as the 3-fund “US stock” bucket)
• VTWO (Russell 2000 mentioned)

International / Emerging

• VXUS (modeled as the international bucket)
• EMBD (used as a data substitute symbol in Testfolio)
• EMB (used as a substitute for missing history of EMBD)
• VWO is referenced in an error-checking context (appears in an LLM prompt output) and is not clearly tied to a specific factor beyond appearing as an ETF label in graphics.

Bonds / Fixed Income

• BSV (short-duration Treasury bond ETF; expense ratio corrected in LLM verification)
• BND (modeled aggregate bond exposure; alternatives are described as sometimes folded into the bond bucket)

Alternatives

• Gold / precious metals: gold exposure is referenced, but a specific ticker is not clearly confirmed in subtitles.
• Silver: referenced, but the ticker is not clearly confirmed in subtitles.
• Managed futures: referenced (managed futures allocation exists in the 16-fund mix).

Cash / T-Bills

• T-bills referenced generally (short-duration bonds/cash-like exposure is part of the portfolio philosophy).

Key Numbers and Performance Metrics

Fee Details

Advisor charges:

• $10,000 per year flat fee
• plus 0.69 basis points (≈ 0.7% AUM fee)
• plus ETF expenses

Net effective burden:

• ~1.6% of invested assets per year (presenter’s estimate/model)

10-Year Performance Comparison (Starting $10,000; Backtest)

Before Fees (gross of the 1.6% advisor fee impact)

• 16-fund portfolio CAGR: 7.97%
• 3-fund portfolio CAGR: 7.43%
• Ending value (as presented): about $21,500 vs ~$20,500 (rounded)
• Volatility / Sharpe: presenter notes the 16-fund had better risk-adjusted metrics, including a higher Sharpe (exact Sharpe value not shown in subtitles).

After Baking in the 1.6% Annual Fee

• 16-fund CAGR drops to: 6.26% (Claude output)
• 3-fund CAGR remains: 7.43% (as presented)
• Sharpe ratio decreases for the 16-fund; drawdowns/gains shift year-to-year (consistent with expected fee drag).

Testfolio Results (Directionally Similar; Alignment/Backfill Assumptions)

• Before fee (approx):
  • 16-fund CAGR: ~8.5%
  • 3-fund CAGR: just under 8%
• After 1.6% fee (“drag” in Testfolio):
  • 16-fund CAGR: 6.87%
  • 3-fund CAGR: just under 8%

Main conclusion:

• A 1.6% annual fee is a big deal, even illustrated using a $10k example over roughly 10 years.

Why the 16-Fund Beat (Before Fees)

The presenter attributes outperformance mainly to:

• Less drawdown in 2022 (help from alternatives + short-duration bonds)
• “Won big” in 2025, driven by commodities (notably small exposure to silver and gold)

Also noted:

• Claude used backfilling because some ETFs launched later (some funds had only ~5–7 years of history).

LLM Data Quality Checks (Methodology Validity)

• The creator compared Claude vs Gemini outputs for expense ratios and found errors.
• Examples corrected:
  • BSV: Claude 0.03% (correct); Gemini showed 0.04%
  • VTWO: discrepancies in reconciliation; presenter states the actual is 0.06%. Claude and Gemini differed (Claude 0.07, Gemini 0.10 in the transcript context).

Takeaway:

• LLMs can be wrong, so inputs should be verified.

Methodology / Step-by-Step Framework Used

• Input extraction

  • Use viewer-provided allocations for a 16-ETF portfolio (Traditional IRA + Roth IRA; dollar amounts not provided).

• AI-based portfolio construction & validation

  • Use Claude (Anthropic) to:
    • generate an asset allocation graphic
    • list expense ratios from tickers
    • iterate prompts (e.g., ask Claude to add expense ratios after initial output)
  • Use Gemini as a cross-check:
    • verify whether expense ratios and ticker descriptions in the graphic are correct
    • identify discrepancies; manually verify BSV and VTWO expense ratios.

• Compare to a 3-fund portfolio

  • Ask AI to map/fold the 16-fund exposures into a canonical set:
    • US stocks → VTI (combines VOO + VTWO)
    • International stocks → VXUS
    • Bonds/alternatives → BND (combines fixed income plus alternatives like managed futures and some metals into a “bond-like” model bucket)
  • Use Claude to estimate 10-year performance (starting $10,000).

• Fee impact modeling

  • Apply a 1.6% effective annual fee to the 16-fund portfolio and rerun comparisons.
  • Repeat in Testfolio (paid version needed for >10 tickers):
    • backfill data where necessary (e.g., EMBD filled with EMB)
    • apply 1.6% drag
    • compare 16-fund vs 3-fund.

Recommendations / Conclusions Stated

• Overall portfolio framing (conceptually):
  • Mostly low-cost index ETFs
  • Some allocation to gold/silver
  • Some managed futures
  • Some short-duration fixed income / T-bills
• Main critique:
  • The 1.6% annual fee burden is too high
  • The fee causes the 16-fund portfolio to underperform the 3-fund setup on a risk-adjusted and return basis (per the model).

Alternative suggestions:

• Consider lower-cost advice (example benchmark: Vanguard advisory services at 0.3%; not presented as endorsement of a specific firm).
• For tax/retirement planning, seek someone who specializes in retirement planning, potentially one-time/flat/hourly rather than AUM-based—especially for:
  • Roth conversion analysis
  • addressing widow’s penalty considerations

Important balance/caution:

• Don’t automatically abandon an advisor solely due to fees—evaluate whether the setup meets your goals and whether you can replicate the planning value elsewhere.

Disclosures / Disclaimers Included

• “I’m not a financial advisor. I’m not a registered investment advisor.”
• Framing as personal perspective and educational analysis (“two cents,” “maybe we’ll have some fun”).

Presenters / Sources

• Presenter/author: The YouTube video speaker (name not provided in subtitles)
• AI tools/sources used:
  • Claude (Anthropic)
  • Gemini (Alphabet/Google)
  • Testfolio (online backtesting tool)