Modeled on a real profile. Names, employers, counts, and placements are altered sample data. Nothing here represents a live Overflow cohort result.
Project-fit profile
Riley Okafor
A three-time software founder who uses coding agents to decide what companies should own and gets those systems into use.
Career context comes from LinkedIn. Current depth comes from observed agent work.
LinkedInSessions01 / Agent footprint
Observed agent activity.
Network placement uses submitted profiles that follow the same rules. Tool mix uses the same dates for both tools.
Anonymous cohort viewCurrent assessment
Your proof, in context.
Compared with the current assessment cohort.
Captured agent activity82nd percentile · upper third · 64/64 reported580agent sessions 40 lowmiddle 50% · 160–7201,200 high
Captured sessions across the agent-tool windows reported in the sample.
Delegation rate76th percentile · upper third · 64/64 reported48%% of scanned sessions 4 lowmiddle 50% · 18–5170 high
Same-source scanned sessions that used subagents.
Parallel-work rate88th percentile · upper third · 64/64 reported74.2%% of active days 3 lowmiddle 50% · 16–5890 high
Observed active days with more than one interactive session.
Owned into real use95th percentile · upper third · 64/64 reported100%% of work examples 0 lowmiddle 50% · 25–75100 high
Work examples that reached real use under this person's ownership.
Independent verification95th percentile · upper third · 64/64 reported100%% of work examples 0 lowmiddle 50% · 20–70100 high
Work examples with a separate retained review or acceptance check.
GitHub active months91st percentile · upper third · 64/64 reported12active months 1 lowmiddle 50% · 5–1112 high
Months with authenticated authored GitHub activity in the sample year.
Each curve uses only profiles that report a compatible version of that metric and appears at eight reported values. Agent-operating curves retain their source-compatible denominators. Baseline activity metrics span older assessment versions. This compares submitted evidence, not career seniority.
signal received · curiosity is a useful metric
Matched-window comparison
44-day shared windowNormalized Claude / Codex mix
Claude Code196 · 4.5/day
Codex341 · 7.8/day
Both counts use 2026-06-11–2026-07-24. 537 session records are inside the shared window. This is session-record share, not time, preference, quality, or completed work. Counts come from local session records. Fixed checks confirm the totals without reading the raw session text.
Assessment02 / Agent operating profile
Acts as the director and reviewer of an agent delivery system.
How this person delegates, coordinates, owns, and checks agent work.
SessionsAssessmentFrame the outcome
Starts with the operating problem, then sets scope and acceptance rules.
Split owned work
Assigns bounded workstreams with explicit ownership and completion tests.
Check independently
Uses a separate review before accepting material implementation results.
Operate and repair
Returns to delivered systems to fix failures and improve behavior.
What is distinctive
Uses coding agents for operating decisions as well as implementation work.
Reusable operating system
Reuses bounded reviewer briefs, source rules, and acceptance checks across projects.
Where the evidence bends
The captured sessions show direction and review, not direct code authorship.
03 / Career & industries
Where their expertise comes from.
Career context establishes duration and responsibility. Observed work shows what is active now.
AI consulting and custom softwareMid-market and private-equity-backed companies
Career contextLed advisory work for mid-market and private-equity-backed companies with responsibility for product and operating decisions.
Observed workThree classified arcs cover delivery systems, commercial workflows, client integrations, and post-launch operation.
LinkedInSessionsEnterprise software and B2B SaaSMobile products and conversation intelligence
Career contextLed enterprise product launches through two acquisitions and sustained commercial ownership.
Observed workCurrent arcs cover CRM, delivery platforms, account reporting, data repair, and a cross-platform subscription product.
LinkedInSessionsCommercial fabrication and interiorsOffices, hospitality, and public spaces
Career contextSix years as founder and chairman of a wood-and-metal fabrication company.
Observed workThe retained agent window has no qualifying fabrication work arc, so current digital depth is not inferred.
LinkedInAssessmentConsumer softwareSubscriptions and cross-platform release
Career contextProduct-company history establishes commercial ownership and adoption responsibility.
Observed workTwo arcs include payments, desktop notarization, mobile test distribution, and consumer application operations.
LinkedInSessions04 / Subject matter
What they know.
Each area names the work that supports it. Generic tool use does not qualify.
Commercial product formation
Connects product definition, pricing, delivery, and adoption during the first operating stage.
LinkedInAgent operating models
Defines worker boundaries, reusable briefs, review gates, spend controls, model routing, and recurring jobs.
SessionsProfessional-services operations
Builds staffing views, review preparation, account reporting, delivery vocabulary, and privacy boundaries.
SessionsEvidence and data integrity
Works on schema compatibility, sync freshness, recovery, privacy limits, and reports that distinguish missing data.
Sessions05 / Matching decision
How to use the profile.
Make the staffing decision without turning uncertainty into a score.
Founder or product operator for owned software
Best for ambiguous company workflows that need a business decision, an agent operating model, and a working internal product.
LinkedInSessionsWhen engineering risk dominates
Pair with dedicated security, database-scale, testing, continuous-integration, or reliability ownership.
AssessmentDirect code authorship
The retained evidence does not establish how much code the person wrote directly.
Assessment07 / Sources & limits
Where the data stops.
Sources retained in this profile: LinkedIn, Sessions, Assessment.
Cross-tool comparison uses only the 44-day shared window because source retention differs.
AssessmentCareer context is self-reported. Captured work remains the source for current depth.
LinkedInAssessmentThe report does not independently verify business outcomes or sole authorship.
AssessmentEvidence index
ev-004LinkedIn
Reviewed career history supports founder, industry, operating, and product-formation context.
ev-003Sessions
Three classified arcs support current delivery, operations, product, workflow, and data-integrity claims.
ev-002Sessions
Observed sessions and tool metadata support the agent workflow, concurrency, subagent, and reusable-review claims.
ev-001Assessment
Deterministic aggregation found 196 Claude and 341 Codex sessions in the exact shared 44-day window.
ev-005Assessment
The retained evidence does not independently prove outcomes, sole authorship, or unobserved current industry work.
Generated 2026-08-02T12:00:00Z