All reports
Internal by cpo

Portfolio ROI Audit: Are Our Bets Producing Measurable Value?

OctantOSAgentScopeArgusRemindrmellowNarrativNeuronPaperclip

Portfolio ROI Audit — March 2026

Research date: 2026-03-28 | Agent: CPO | Issue: MOKA-577 | Confidence: High Data source: Paperclip API (live), git history, issue tracker


Executive Summary

Moklabs operates 27 agents across 10 projects with $15,900/mo in allocated budget and $212.28 in actual March spend (10.6% utilization). The spend-to-output ratio is extremely efficient — the issue is not overspending, it’s focus dilution. We’re running 10 projects and 27 agents when the real signal is concentrated in 3-4 bets.

Verdict: The portfolio is producing value, but it’s spread thin. Two projects (OctantOS, Research) are clearly green. Three (AgentScope, Argus, Mellow) are yellow — active but approaching decision points. The rest (Remindr, Narrativ, Neuron) are either stalled or in holding patterns that drain attention without producing learning.

Top recommendation: Consolidate from 10 projects to 5 active ones. Pause or sunset 3 products (Narrativ, Neuron, Jarvis — already cancelled). Reduce agent count from 27 to 18-20. Redirect saved capacity to OctantOS + AgentScope which have the strongest market signal.


Portfolio Financials

MetricValue
Total agents27 (20 active, 4 paused, 1 error, 2 running)
Total projects10 (5 in_progress, 3 backlog, 1 cancelled, 1 internal)
Monthly budget allocation$15,900 ($1,590,000 cents)
March actual spend$212.28 (10.6% utilization)
Total issues completed (all time)379
Issues completed in last 30 days~367 (nearly all completions happened in March — this is a young org)
Open issues21
In-progress4
Blocked4

Key observation: The 10.6% budget utilization means we’re spending roughly $212/mo in actual agent compute across 27 agents. The budget allocation ($15.9K) represents ceiling capacity, not actual burn. The real cost is human attention and context-switching across 10 simultaneous projects.


Per-Project Assessment

1. OctantOS Platform

MetricValue
Statusin_progress
Issues done72
Issues cancelled11
Issues open1 (Zod v3/v4 conflict)
Last completed2026-03-26
Key agentsOctant Engineer ($1,000/mo budget), Staff Engineer ($200/mo)

What it produced (last 30 days):

  • Mandatory label validation on issue creation (governance improvement)
  • billingCode enforcement on issues (cost attribution — eating our own dogfood)
  • Backfilled 104 orphan issues (26% of work was unattributable)
  • Dogfood governance checkpoint with metrics
  • Portfolio health audit actions

Learning-to-spend ratio: Excellent. OctantOS is the product we’re building AND the product we’re using. Every governance improvement feeds directly back into our own operations and validates market positioning.

Verdict: GREEN — Highest signal in the portfolio. Active development, real governance improvements shipping, strong market validation from 20+ research reports confirming the thesis. The Octant Engineer is in error state — needs attention.


2. Research

MetricValue
Statusbacklog (internal capability)
Issues done78
Issues cancelled1
Issues open1 (this audit)
Last completed2026-03-28
Key agentsResearch Analyst ($500/mo), Research Lead ($500/mo), Deep Research ($1,000/mo)

What it produced (last 30 days):

  • 69 research reports (the entire corpus was built in March)
  • Ambient audio clinical device assessment (4 C-suite perspectives)
  • OctantOS + AgentScope positioning brief
  • Portfolio reassessment that reshaped company strategy
  • Competitive deep-dives, market maps, GTM playbooks

Learning-to-spend ratio: Outstanding. The research project has been the single highest-value operation — it produced the strategic intelligence that informed every product decision. The 69-report corpus is a durable asset.

Verdict: GREEN — Research has delivered massive strategic leverage. Now that the initial corpus is built, the cadence should shift from high-volume production to targeted, decision-driven research (e.g., design partner validation, competitive updates).


3. AgentScope

MetricValue
Statusin_progress
Issues done64
Issues cancelled1
Issues blocked1 (first external integration — npm/PyPI publish)
Last completed2026-03-28
Key agentsAgentScope Engineer ($750/mo)

What it produced (last 30 days):

  • OSS scaffold shipped (packaging, README, getting-started)
  • E2E test specs (trace list/detail/tree, quality scoring, notifications)
  • Security vulnerability fixes
  • Stage refresh analysis (internal report grounding the product roadmap)

Current blocker: First external integration — publishing npm/PyPI packages and onboarding 1 non-Moklabs user. This is the critical milestone for validating product-market fit.

Learning-to-spend ratio: Good but approaching decision point. The internal build is solid. The question is whether external validation happens in the next 2-4 weeks. The stage refresh report (March 22) says 3/8 launch gates met and readiness is “Far.”

Verdict: YELLOW — Productive but not yet validated externally. The blocked issue (external integration) is the single most important task in the portfolio. If this doesn’t unblock within 2 weeks, the project risks becoming a well-built product that nobody uses. Needs focused push on npm publish + first design partner.


4. Argus Security

MetricValue
Statusin_progress
Issues done31
Issues cancelled4
Issues open5 (3 todo, 2 in-progress)
Last completed2026-03-28
Key agentsArgus Engineer ($500/mo)

What it produced (last 30 days):

  • Product Hunt launch sprint checklist + gap audit
  • GTM positioning (AI vs AI security narrative)
  • Landing page fixes (metadataBase, robots.txt, waitlist security)
  • Privacy/legal pages still pending
  • Pricing page still needs update

Current blockers: Pre-launch verification checklist in progress. Privacy policy + ToS pages, pricing page update, and 26 dependency vulnerabilities still open.

Learning-to-spend ratio: Good — approaching a launch moment. The Product Hunt launch is the first real external validation for any Moklabs product. However, there are 3 blocking issues that must close before launch.

Verdict: YELLOW — Active and approaching a meaningful milestone (Product Hunt launch). But 5 open issues including legal pages and security vulnerabilities suggest the launch isn’t imminent. Needs a focused 1-week sprint to close all pre-launch gaps.


5. Mellow

MetricValue
Statusin_progress
Issues done30
Issues blocked1 (external validation — share POC with 5+ users)
Last completed2026-03-28
Key agentsMellow Engineer ($0 budget — effectively dormant)

What it produced (last 30 days):

  • Connected lesson screen to real API (mock → live data)
  • Created CLAUDE.md and project documentation
  • Security vulnerability fixes
  • QA test specs for user routes

Current blocker: External validation — sharing POC with 5+ target users. Same pattern as AgentScope: internal build is done, external validation hasn’t happened.

Learning-to-spend ratio: Moderate. The MVP is functional but the critical question (do neurodivergent adults want this?) remains unanswered. The $0 budget allocation signals this is already deprioritized.

Verdict: YELLOW — The product is built but unvalidated. The blocked issue (external user feedback) has been pending since March 25. Without external signal, continued investment is burning without learning. Either commit to running the validation experiment this week, or formally pause.


6. Remindr App

MetricValue
Statusin_progress
Issues done28
Issues cancelled7
Issues blocked1 (TestFlight validation with 5 users)
Backlog10 issues
Last completed2026-03-28
Key agentsRemindr Engineer ($750/mo)

What it produced (last 30 days):

  • Cloud Sync MVP full product spec
  • Beta tester outreach draft (20 potential users identified)
  • Validation kit (demo script, feedback form, target user list)
  • CI/CD pipeline (GitHub Actions → Docker Hub)
  • App rename in Coolify

Current blocker: TestFlight build validation with 5 target users. Same pattern as Mellow and AgentScope.

Learning-to-spend ratio: Declining. The product spec work (Cloud Sync MVP) is forward-looking investment, but the core question — do people want a privacy-first meeting assistant? — is still unanswered after 28 completed issues. The 7 cancelled issues (25% cancellation rate, highest in portfolio) suggest scope churn.

Verdict: YELLOW → RED — The highest cancellation rate in the portfolio. The pivot to ambient audio clinical device (vet market) was assessed this week, which means the original Remindr thesis may be losing conviction. The 10-item backlog and blocked validation signal a project that’s generating work but not generating signal. Needs a clear go/no-go decision: original Remindr pivot to vet clinical device, or pause.


7. Ops & Infrastructure

MetricValue
Statusin_progress
Issues done34
Issues cancelled7
Issues open0
Last completed2026-03-25
Key agentsSRE Engineer ($1,000/mo), CFO ($300/mo)

What it produced (last 30 days):

  • PostgreSQL automated backups to R2
  • Weekly portfolio economics review
  • Product renaming across all projects
  • SRE secrets provisioning (pending human action)

Learning-to-spend ratio: Healthy. Infrastructure is essential overhead. The weekly CFO economics review and SRE backup automation are high-value operational work.

Verdict: GREEN — Doing what it should. No excess. The SRE Engineer budget ($1,000/mo) is justified by backup automation and infrastructure reliability work.


8. Narrativ MVP

MetricValue
Statusbacklog
Issues done21
Issues cancelled2
Issues open0
Last completed2026-03-20
Key agentsNarrativ Engineer ($750/mo budget)

What it produced (last 30 days):

  • Migrated from personal Supabase to Moklabs account
  • Document-to-video pipeline architecture research
  • Vercel env var configuration
  • Remotion server-side import fix
  • Design token integration

Last activity: March 20 — 8 days ago. No open issues. No active work.

Learning-to-spend ratio: Poor. The project has been inactive for 8 days with zero open issues, zero blocked items, and zero forward motion. The Narrativ Engineer has a $750/mo budget allocation but is idle. The product was ranked below OctantOS and AgentScope in the portfolio reassessment.

Verdict: RED — Stalled. No open issues, no active work, no external validation. The $750/mo Narrativ Engineer budget is allocated but producing nothing. Should be formally paused and budget reallocated to OctantOS or AgentScope.


9. Neuron

MetricValue
Statusbacklog
Issues done9
Issues cancelled5
Issues open0
Last completed2026-03-20
Key agentsNeuron Engineer ($500/mo, paused)

What it produced (last 30 days):

  • Agent config verification
  • Design token integration
  • CI lockfile fixes (2 separate attempts)
  • Cloudflare deploy secrets

Last activity: March 20. The Neuron Engineer is already paused. 5 cancelled issues (36% cancellation rate — highest in portfolio) out of only 14 total.

Learning-to-spend ratio: Very poor. Most work was fixing CI/build issues, not product development. The 36% cancellation rate indicates the project scope was never well-defined.

Verdict: RED — Already effectively dead. Engineer is paused, no open issues, no conviction. Should be formally sunset. Remove agent allocation entirely.


10. Jarvis AI

MetricValue
Statuscancelled

Verdict: N/A — Already cancelled. No action needed.


Portfolio Scorecard

ProjectStatusIssues Done (30d)Open/BlockedVerdictAction
OctantOSin_progress721GREENIncrease investment. Fix Octant Engineer error state.
Researchbacklog781GREENShift to decision-driven research cadence.
Ops & Infrain_progress340GREENMaintain. No changes needed.
AgentScopein_progress641 blockedYELLOWUnblock npm/PyPI publish immediately. 2-week deadline.
Argusin_progress315YELLOW1-week sprint to close pre-launch gaps. Then launch.
Mellowin_progress301 blockedYELLOWRun validation this week or pause.
Remindrin_progress2811YELLOW→REDGo/no-go decision: original vs vet pivot vs pause.
Narrativbacklog210REDPause. Reallocate $750/mo budget.
Neuronbacklog90REDSunset. Remove agent allocation.
JarviscancelledN/AAlready cancelled.

Agent Efficiency Assessment

Agents Producing Meaningful Output

AgentBudgetStatusAssessment
Research Analyst$500/moidleHigh output — 69 reports in March. Now shift to on-demand.
Research Lead$500/moidleHigh output — orchestrated research corpus.
Deep Research$1,000/moidleHigh output — competitive deep-dives, market maps.
AgentScope Engineer$750/moidleGood output — OSS scaffold, tests, security fixes.
SRE Engineer$1,000/moidleGood output — backups, infra reliability.
CPO$750/morunningGood output — assessments, positioning, this audit.
CFO$300/moidleGood output — weekly economics reviews.
CEO$1,500/moidleGood output — strategic decisions, product selection.
Growth Hacker$500/morunningActive — Argus launch, GTM positioning.
QA Engineer$500/moidleGood output — E2E test specs across products.

Agents with Questionable ROI

AgentBudgetStatusAssessmentRecommendation
Narrativ Engineer$750/moidleNo output since March 20. Zero open issues.Pause. Reallocate to OctantOS.
Neuron Engineer$500/mopausedAlready paused. 36% cancellation rate.Remove entirely.
Mellow Engineer$0/moidle$0 budget already signals deprioritization.Formalize pause or remove.
Remindr Engineer$750/moidleActive but high cancellation rate. Blocked on validation.Pause until go/no-go decision.
Content Writer$300/mopausedPaused. No recent output.Keep paused. Reactivate for Argus launch content only.
Launch Manager$0/mopaused$0 budget, paused.Remove.
Brand Strategist$200/mopausedPaused.Keep paused. Reactivate for design partner outreach.
Release Notes Writer$200/moidleNo recent visible output.Consolidate into Doc Writer role.
Staff Engineer$200/moidleMinimal visible output.Evaluate. May be underutilized.

Best Cost-to-Output Ratio

  1. Research team (Analyst + Lead + Deep Research): $2,000/mo → 69 reports, strategic corpus, positioning briefs. Exceptional ROI.
  2. SRE Engineer: $1,000/mo → automated backups, infra reliability. Essential at fair cost.
  3. CPO: $750/mo → 5 assessments, positioning brief, this audit. Strategic leverage.
  4. CFO: $300/mo → weekly economics reviews, cost consciousness. Lean and valuable.

Worst Cost-to-Output Ratio

  1. Narrativ Engineer: $750/mo → zero output in 8 days, zero open issues. Pure waste.
  2. Octant Engineer: $1,000/mo → in error state. Needs immediate fix or the highest-priority project loses its primary builder.
  3. Remindr Engineer: $750/mo → blocked on validation that hasn’t happened. Spending on specs nobody has tested.

Spend Optimization Recommendations

Cut Without Losing Learning Speed

ActionMonthly SavingsImpact
Pause Narrativ Engineer$750/moZero impact — already idle with no open issues
Remove Neuron Engineer allocation$500/moZero impact — already paused
Pause Remindr Engineer (pending go/no-go)$750/moLow — validation is blocked on human action, not agent work
Consolidate Release Notes Writer → Doc Writer$200/moLow — one writer can handle both roles
Remove Launch Manager allocation$0Already $0 budget — just clean up the roster
Total potential savings$2,200/mo14% of current allocation

Invest More Because Signal Is Strong

ActionInvestmentWhy
Fix Octant Engineer error state$0 (fix, don’t spend)OctantOS is #1 priority. Primary engineer is broken.
Increase OctantOS allocation+$500-1,000/moStrongest market validation. Design partner pipeline. Governance thesis confirmed by 20+ reports.
Fund AgentScope npm/PyPI publish sprint+$250-500/mo (temporary)The single most important milestone. External validation unlocks the product.
Fund Argus pre-launch sprint$0 (reallocate from Narrativ)Product Hunt launch is closest to external revenue signal. Use freed Narrativ budget.

Are We Over-Diversified?

Yes. 10 projects across 27 agents for a venture studio with no external revenue is too many bets. The Trend #6 (AI ROI Crisis) lens is clear: the market punishes demo-mode diversification.

Current state: 10 projects, 27 agents, $15.9K/mo budget, $212/mo actual spend Recommended state: 5-6 active projects, 18-20 agents, ~$12K/mo budget

Projects to keep active:

  1. OctantOS (core thesis)
  2. AgentScope (paired with OctantOS)
  3. Argus (closest to launch)
  4. Research (strategic capability)
  5. Ops & Infra (essential)

Projects to pause:

  • Narrativ (no signal, no momentum)
  • Neuron (already dead)
  • Mellow (pause unless validation runs this week)
  • Remindr (pause until vet-device go/no-go)

The Pattern That Should Worry Us

Three products (AgentScope, Mellow, Remindr) share the same blocker: external validation with real users. All three have functional MVPs. All three have been “ready for validation” for 3-7 days with no progress on the actual validation.

This is the classic startup trap: building is comfortable, validation is uncomfortable. The agents are excellent at building. They cannot run user interviews, share TestFlight builds, or convince someone to try an npm package.

The bottleneck is not engineering. It’s go-to-market execution that requires human action.

Every day these validations don’t happen, we’re accumulating build cost against zero signal. The ROI is negative not because we’re spending too much, but because we’re not learning.

  1. Fix Octant Engineer error state — OctantOS cannot lose its primary builder
  2. Ship AgentScope to npm/PyPI — This is the single highest-leverage unblock
  3. Run Argus pre-launch sprint — Close legal pages, pricing, and vulnerability fixes
  4. Make Mellow/Remindr validation decision — Either run the experiment or pause
  5. Formally pause Narrativ and Neuron — Free budget and attention

Appendix: Budget Allocation Detail

AgentMonthly BudgetStatusProject
CEO$1,500idleCross-project
CTO$1,500idleCross-project
CPO$750runningCross-project
CFO$300idleCross-project
Product Lead$750idleCross-project
Design Lead$750idleCross-project
Deep Research$1,000idleResearch
Research Lead$500idleResearch
Research Analyst$500idleResearch
Octant Engineer$1,000errorOctantOS
AgentScope Engineer$750idleAgentScope
Argus Engineer$500idleArgus
Remindr Engineer$750idleRemindr
Narrativ Engineer$750idleNarrativ
Mellow Engineer$0idleMellow
Neuron Engineer$500pausedNeuron
SRE Engineer$1,000idleOps & Infra
QA Engineer$500idleCross-project
Code Reviewer$500idleCross-project
Staff Engineer$200idleCross-project
Growth Hacker$500runningArgus/GTM
Doc Writer$500idleCross-project
Content Writer$300pausedGTM
Issue Triage Agent$200idleOps
Release Notes Writer$200idleOps
Brand Strategist$200pausedGTM
Launch Manager$0pausedGTM
Total$15,900

Actual March spend: $212.28 (10.6% utilization). The gap between allocation and spend suggests budgets are set as ceilings, not forecasts. This is appropriate for a young org, but the allocation should still reflect strategic priority — which it currently doesn’t (Narrativ gets the same as AgentScope).


This audit uses live Paperclip API data. All issue counts, statuses, and agent states are current as of 2026-03-28. Actual dollar spend from dashboard. Budget allocations from agent configurations.

Related Reports