Messy Fund Update

Archived snapshot

Fund / Signal update · 15 August 2026 · vnext-2026-07-24

Weekly workflow proof for Guru micro test funds on Base (base01, base02, base04, base05, base06). Screening prepares evidence; the Chair decides; the system records the resolution; only then can execution mutate the book. Macro and narrative below are the postures those sessions used, not a live market refresh at publish time.

Custody / provider

These Messy-managed books use the Guru Lotus vault adapter on Base for custody and share accounting. Messy’s engine runs research, council decisioning, and trade construction; capital sits in Guru’s vault contracts. Our trading key can trade but cannot withdraw, a limit enforced by the vault contracts, not our software. Provider-side contract failure is venue risk. Keys live in Turnkey; our backend never holds key material.

For the blast-radius map (engine vs custody vs token), see Custody Broke. The Engine Didn’t..

This week at a glance

  • 30 council sessions across 5 micro funds — 21 executed, 4 holds.
  • Week regime path: Risk-off (R-) → Neutral (N) — Effective macro regime is N with fresh coverage and 80% confidence.
  • Narrative tape (session-frozen): Across 16 active narrative(s), 30d momentum vs TOTAL3ES ranges from SocialFi at +30.1pp to NFT at -20.9pp.
  • base02 NAV $36.8 (-12.6% 7d, 5 exec / 0 hold).
  • base01 NAV $89.66 (-6.8% 7d, 5 exec / 0 hold).
  • COOKIE led aggregate screens this week — a candidate signal, not an automatic trade.

What's new in the app

  • Fund pages in the app now include full council minutes — open any fund and tap Council to read what the portfolio committee debated, what the Chair decided, and whether trades were executed. How the AI investment council works →
  • We publish five Guru micro test funds (base01, base02, base04, base05, base06). base04 and base05 joined on 10 July; base06 began council meetings in the week of 1 August. Every change on those test funds goes through screening, council review, and signed execution — the same path future AI-managed funds will use.
  • This week we specialized the screening playbooks on base04, base05, and base06: meme-nano, asymmetric-micro, and liquid-control. base01 and base02 keep the shared momentum + social templates. Screens still stop at ideas — council and execution decide size, venue, and fills. Rolling 7-day aggregates can mix old and new context until the new playbooks have a full week of runs.

Week regime · what the councils used

Regime path

Risk-off (R-) → Neutral (N)

Mon R- exec
Tue R- exec
Wed R- exec
Thu N exec
Fri N target
Sat N exec

Sampled from base01 sessions across the week

Closing macro brief

Neutral (N) · as of 2026-08-15

Effective macro regime is N with fresh coverage and 80% confidence.

Soft narrative tape

Across 16 active narrative(s), 30d momentum vs TOTAL3ES ranges from SocialFi at +30.1pp to NFT at -20.9pp.

  • 100.0% of beta book value (attributed by primary narrative) sits in narratives with negative trailing 30d momentum vs TOTAL3ES.

Flagged in base06 Chair notes

Micro funds · week scan

Fund NAV 7d Sessions Exec Hold
base01 $89.66 -6.8% 9 5 0 Council
base02 $36.8 -12.6% 5 5 0 Council
base04 $40.26 -4.3% 5 5 0 Council
base05 $46.93 +0.6% 6 4 1 Council
base06 $23.97 +1.1% 5 2 3 Council

base01 — council & book

Closing session · executed reconciled · 15 August 2026

The rebalance executed and reconciled, trimming Mrdn, exiting BID, and adding TOBY, ACU, and UP while holding TIBBIR, within a fund posture ratified at roughly cash 27%, base 40%, and high beta 33%; realized post-execution weights are not yet available to confirm against that target.

Week strip: Mon exec · Tue exec · Wed exec · Thu exec · Fri target · Sat exec · 9 sessions

Open Council minutes →

Mon exec
Tue exec
Wed exec ×4
Thu exec
Fri target
Sat exec

Week net · Cash 27.2% → 27% · Base 33.2% → 40% · High-beta 39.5% → 33%

Tue exec flay carries 32.6% of a two-position sleeve — outsized single-name concentration that a same-size reduce/trim only partially addresses.

Closing session

Cash 26.7% → 27% · Base 40.6% → 40% · High-beta 32.7% → 33%

Recommended movement is cash +0.3%, base -0.6%, high beta +0.4%; each change remains inside the configured session movement caps of cash 5.0%, base 5.0%, and high beta 5.0%.

  • Book concentration (top-1 18.1% NAV, effective N 2.45) sits below the 2-10 breadth band the sleeve target supports; this recurs each session until addressed and is not itself a same-day reversal concern.
  • tibbir RSI-short at 75.4 is entry-timing information only per the measured-axes study; not a basis to trim a name whose fresh/aggregate rank (2/2) still supports the thesis.
  • PM seats disagree on mrdn: exit_strategist and macro_tactical favor a trim on coverage-gap/concentration grounds while others hold it in position_intent unchanged \u2014 sizing of the trim matters more than…
Current book
Cash 27.1% · Base 40.2% · High-beta 32.7%
CASH (USDC) 27.1% BASE (WETH) 40.2% MRDN 9.7% TIBBIR 6.8% ACU 5.5% UP 5.5% TOBY 5.2%

Top aggregate candidates — 7-day lookback , as of 15 August 2026

#1 COOKIE Appeared in 7/7 source run(s) with avg rank 3.00, best rank 1, recency score 29.73.
#2 TIBBIR Appeared in 5/7 source run(s) with avg rank 2.40, best rank 1, recency score 28.56.
#3 WOLF Appeared in 5/7 source run(s) with avg rank 7.20, best rank 5, recency score 9.36.

base02 — council & book

Closing session · executed reconciled · 14 August 2026

The council's rebalance executed and reconciled this session, reducing flay, increasing mrdn, and trimming rwa.

Week strip: Mon exec · Tue exec · Wed exec · Thu exec · Fri exec · 5 sessions

Open Council minutes →

Mon exec
Tue exec
Wed exec
Thu exec
Fri exec

Week net · Cash 26.9% → 29.3% · Base 33.9% → 31.7% · High-beta 39.1%

Tue exec Mrdn's flagged +763bps route gap is a real execution ceiling on any full exit; size or stage the sell accordingly.

Closing session

Cash 29.3% · Base 31.7% · High-beta 39.1%

No bucket movement is recommended because current allocation already satisfies the computed posture.

  • rwa's RelStr collapse (-24) and disappearance from the fresh screen is one session of data; guidance requires sustained deterioration across several sessions or a hard trigger before treating it as an exit…
  • flay was reduced last session; a further reduction now needs a named thesis change beyond aggregate absence, which has not been clearly stated by the proposing PM.
  • avnt has zero aggregate history, no measured ATR/dist-to-high/RSI data, and only fresh-run conviction — appropriate for advisory sizing caution only, not a block.
Current book
Cash 29.8% · Base 32.3% · High-beta 38%
CASH (USDC) 29.8% BASE (WETH) 32.3% MRDN 18.4% COOKIE 12.3% RWA 4.0% FLAY 3.3%

Top aggregate candidates — 7-day lookback , as of 15 August 2026

#1 COOKIE Appeared in 7/7 source run(s) with avg rank 1.57, best rank 1, recency score 36.34.
#2 WOLF Appeared in 6/7 source run(s) with avg rank 3.83, best rank 3, recency score 25.93.
#3 MRDN Appeared in 5/7 source run(s) with avg rank 4.60, best rank 2, recency score 13.42.

base04 — council & book

Closing session · executed reconciled · 14 August 2026

The rebalance executed and reconciled: cocoro was exited and taboshi entered with toby and apu trimmed, while the wolf leg was left as an unresolved residual due to a missing execution price.

Week strip: Mon exec · Tue exec · Wed exec · Thu exec · Fri exec · 5 sessions

Open Council minutes →

Mon exec
Tue exec
Wed exec
Thu exec
Fri exec

Week net · Cash 28.9% → 27.7% · Base 34.5% → 36.6% · High-beta 36.6% → 35.6%

Tue exec 100% of high-beta book value sits in narratives (Meme-primary) with negative 30d momentum vs TOTAL3ES — a structural narrative headwind, not a single-token issue.

Closing session

Cash 27.7% · Base 36.6% · High-beta 35.6%

No bucket movement is recommended because current allocation already satisfies the computed posture.

  • Sleeve is fully invested; any new entry must be funded by a recorded reduction/exit within mvs-base04-1, not by uncommitted capacity.
  • Cocoro's coverage has gone dark (absent from latest fresh run, only 3/7 aggregate appearances) — directionally supports exit but paper trail is thin, size the confidence accordingly.
  • Apu's route gap is flagged at +2404bps; an exit there would consume most of the position's value in execution cost and is not economically supportable this session.
Current book
Cash 28.1% · Base 44.6% · High-beta 27.3%
CASH (USDC) 28.1% BASE (WETH) 44.6% TOBY 10.6% APU 9.4% TABOSHI 7.3%

Top aggregate candidates — 7-day lookback , as of 15 August 2026

#1 TOBY Appeared in 7/7 source run(s) with avg rank 3.71, best rank 2, recency score 30.40.
#2 WOLF Appeared in 6/7 source run(s) with avg rank 3.00, best rank 1, recency score 28.19.
#3 APU Appeared in 6/7 source run(s) with avg rank 5.17, best rank 2, recency score 15.67.

base05 — council & book

Closing session · executed reconciled · 14 August 2026

The council's rebalance executed and reconciled, trimming jarvis and adding taboshi within an unchanged fund posture of roughly 28% cash, 35% base, and 37% high beta.

Week strip: Mon exec · Tue hold · Wed exec · Thu exec · Fri exec · 6 sessions

Open Council minutes →

Mon exec
Tue hold
Wed exec
Thu exec ×2
Fri exec

Week net · Cash 28.2% → 27.7% · Base 34.1% → 34.9% · High-beta 37.7% → 37.3%

Tue hold Dickbutt was already reduced last session; a full exit today compounds execution friction within roughly one day unless the deterioration case (coverage loss + multi-session score decline) is explicitly the…

Closing session

Cash 27.7% · Base 34.9% · High-beta 37.3%

No bucket movement is recommended because current allocation already satisfies the computed posture.

  • Taboshi ATR% 24h is 12.9%, meaningfully higher than jarvis (3.1%) or up (5.0%) — sizing should reflect this volatility differential.
  • 856-up trend_momentum add has no aggregate screening history depth comparable to taboshi (only 2/7 runs) — thinner durability basis.
  • Jarvis trim funds multiple candidate adds across seats; council must ensure decomposition is sell jarvis→reserve, then reserve→buy legs, not a direct swap framing.
Current book
Cash 27.8% · Base 35% · High-beta 37.2%
CASH (USDC) 27.8% BASE (WETH) 35.0% UP 17.0% TABOSHI 12.5% JARVIS 7.7%

Top aggregate candidates — 7-day lookback , as of 15 August 2026

#1 TABOSHI Appeared in 5/7 source run(s) with avg rank 2.20, best rank 1, recency score 30.08.
#2 JARVIS Appeared in 5/7 source run(s) with avg rank 4.40, best rank 1, recency score 11.60.
#3 UP Appeared in 4/7 source run(s) with avg rank 1.50, best rank 1, recency score 27.44.

base06 — council & book

Closing session · executed reconciled · 14 August 2026

The rebalance executed and reconciled, adding a small link position funded by trimming omi while tbtc, wbtc, and tibbir held unchanged.

Week strip: Mon hold · Tue hold · Wed hold · Thu exec · Fri exec · 5 sessions

Open Council minutes →

Mon hold
Tue hold
Wed hold
Thu exec
Fri exec

Week net · Cash 35.7% → 31.4% · Base 39% → 38.8% · High-beta 25.3% → 29.8%

Tue hold Capacity is compressed: only $1.18 unfunded deployable and one net new-entry slot; any addition beyond the reserve-funded omi top-up competes for the same thin capacity.

Closing session

Cash 31.4% · Base 38.8% · High-beta 29.8%

No bucket movement is recommended because current allocation already satisfies the computed posture.

  • Sell-fundable capacity is only $7.03 against a fully invested book; any new entry (link, ethfi) must be funded via an explicit incumbent reduction/exit decomposed into sell-to-reserve then reserve-to-buy…
  • Macro regime is neutral with mixed EMA momentum (7d vs 3d -1.89, 30d vs 7d -0.91) and a soft crypto-structural undertone (negative stablecoin/TOTAL3 flows, negative BTC ETF flows) — this argues for caution…
  • 100% of beta book value sits in narratives with negative 30d momentum vs TOTAL3ES; this is context, not itself a sell trigger for any single holding.
Current book
Cash 30.9% · Base 38.2% · High-beta 30.9%
CASH (USDC) 30.9% BASE (WETH) 38.2% TBTC 9.1% WBTC 8.1% TIBBIR 7.1% LINK 5.5% OMI 1.2%

Top aggregate candidates — 7-day lookback , as of 15 August 2026

#1 OMI Appeared in 7/7 source run(s) with avg rank 4.00, best rank 1, recency score 24.40.
#2 TIBBIR Appeared in 5/7 source run(s) with avg rank 1.20, best rank 1, recency score 32.81.
#3 LINK Appeared in 5/7 source run(s) with avg rank 3.40, best rank 2, recency score 26.61.

Aggregate signal · Risk reject

✅ Aggregate signal
COOKIE
Cookie DAO
Appeared in 7/7 daily runs across base01 and base02 over the 7-day lookback. Appeared in 7/7 source run(s) with avg rank 3.00, best rank 1, recency score 29.73.
❌ Risk reject
base01
base01 · base01
Rebalance executed and reconciled.

What this teaches the future Fund Agent

1
Macro gates exposure. Use the session-frozen regime the council actually saw (this week: Neutral (N)) — not a later live refresh.
2
Daily screen ≠ automatic trade. Aggregate leaders inform the slate; they do not bypass Chair decision or execution gates.
3
Chair decides, system records. Seats analyze and propose; the Chair adjudicates; minutes and execution receipts stay public on each fund’s Council tab.

Inspect the same workflow

Archive