8 / 8

Product & Strategy 8 projects

Other collectionsBuilt 3Claude in Daily Life 2Strategy Decks 2PRD Library 5

Proposed screens · sample values

Product case study

LinkedIn Priority Applications

Applying is free, so intent is noise.

My own exploration of a LinkedIn problem: an application no longer shows who really wants the job. A scarce, refundable Priority Application evolves Top Choice, with five features, three wireframes, a metrics pyramid, a scenario business case and an experiment.

  • Product Strategy
  • Marketplace
  • Trust and Signal
  • Experiment Design
  • Business Case
personas
5
features
5
wireframes
3
break-even uplift
0.13%
8 sections · 3 original artifacts

Application signal collapseApplying costs seekers nothing, so application volume is a poor proxy for intent. Hirers sift through noise, seekers hear nothing back, and the marketplace promise of the right number of quality applicants, at the right speed and price, weakens. Microsoft's FY26 10-K reports LinkedIn revenue growth of 11% against 41% for Azure, and its FY27 restatement folds LinkedIn's parts into other lines, so LinkedIn is no longer reported as a standalone unit.

Pressing problems

IssueDelightHard to copyMargin movingPriorityWhy this ranking
Applications are a free, noisy intent signalHighHighHighP1Acts on the root cause and helps both sides of the marketplace. Relies on the identity graph a competitor cannot clone.
Agent access is policed by detection, with no sanctioned pathMediumHighMediumP2Defensible because it is identity-bound, but revenue is speculative and it cuts against LinkedIn's anti-scraping stance. Needs research to size.
AI interviews run on a small minority of jobsMediumMediumHighP2A clear hirer upsell and a source of verified intent for the P1 fix, though other firms are adding AI hiring loops too.
Early-applicant advantage rewards speed over fitMediumLowMediumP2A mechanism of the root cause, and the evidence is a single statement. Best addressed once volume has been reshaped.
Match scores give no path to improveHighMediumMediumP3Seekers like it, but it risks keyword gaming and has the weakest tie to hirer-side revenue.

The full matrix: the evidence and the proposed product change behind each score, and how the DHM model is read.

Download the prioritization matrix (PDF) (opens in a new tab)The Word file

What I am solvingThe P1 only: applications are a free, noisy intent signal. It acts on the root cause, helps both sides and rests on the identity graph a competitor cannot clone. The rest are parked. A verified agent tier stays a later option. AI interviews reach the priority lane as the AI-screen filter and a reason for the label. The early-applicant advantage is partly handled by the fit-and-intent sort in F3. The match coach has the weakest tie to hirer revenue.

Hirer side

Finding the real candidates

Share of applicants advancing to review, time to first qualified applicant, cost per qualified applicant.

Seeker side

Hearing back

Response rate per application, applications per hire.

Platform side

No collateral damage

False-positive restriction rate for legitimate users.

The loop

Figure 1, Why the problem compounds: a reinforcing loop of six steps. Applying is free (LinkedIn Apply, AI auto-apply). Volume rises (about 9,600 applications a minute). The count of 100+ carries no signal, so recruiters skim. Most applicants hear nothing. Seekers conclude odds are low. They apply to more jobs to compensate, which returns to applying being free. Centre label: Reinforcing loop, the intent signal degrades with every turn.
Reinforcing loopWhy the problem compounds, from the PRD. 9,600 applications a minute is the midpoint of two sources that disagree by about a third.

Why it matters

Hirer willingness to pay

Hirers pay for access to qualified candidates. If review time per hire rises, a posting is worth less.

Growth under scrutiny

Microsoft's FY26 10-K says LinkedIn revenue increased 11% against 41% for Azure. Its FY27 restatement folds LinkedIn's parts into other lines, so LinkedIn is no longer reported as its own unit.

Seeker trust

Silence after applying pushes seekers toward more volume, which feeds the loop. An answered signal gives a reason to be selective.

Defensibility

A job board can copy a count. It cannot copy a signal built on verified identity, the member graph and engagement history.

Top Choice today

Top Choice todayLinkedIn's Help page (printed 2026-10-30, last updated two years ago) calls it a Premium feature for a select group: a limited number of Easy Apply jobs are eligible, you can mark up to three jobs a month, credits refresh on the first of each month at 12 AM UTC and do not roll over, the poster sees the job marked as a top choice, and you get an application-viewed email when a hirer reviews it. HR Dive reported the October 2023 rollout and LinkedIn's claim that users were on average 43% more likely to get a message back; that figure is as reported, not verified here.

From LinkedIn Help

Stays the same

  • Marked during Easy Apply, with a required 20 to 400 character message to the employer
  • Premium gets 3 a month; credits refresh on the 1st at 12 AM UTC and do not roll over
  • The hirer sees the job was marked a top choice; the seeker gets an application-viewed email

Proposals, not facts

Proposed changes

  • Free members get 1 priority slot a month (today Premium only), refunded on withdrawal or if never opened
  • Priority or Standard is an explicit choice, with the slots left, instead of a tick box
  • The job page shows tiered applicant counts and the hirer reply rate, replacing "Over 100 applicants"
  • Hirers get a priority lane with filters, a 7-day reply expectation and a fit-and-intent default sort
  • The message and the reason for the label show in the lane; links and contact details are not allowed

Where it falls short

Premium only

Seekers on the free plan cannot show intent at all.

Three a month

Up to three jobs a month, so it flags a handful of applications among hundreds of millions.

A limited set of jobs

Only a limited number of Easy Apply jobs are eligible for it.

Not used to rank

The hirer sees that the job was marked. It does not count, sort or route applications.

Trends

In hiring marketplaces

AI auto-apply

Volume keeps rising

Agents, and LinkedIn's own Premium Apply Assistant, apply on a seeker's behalf. Applications are reported up 45% a year (NYT, July 2026), so count-based signals keep decaying.

Friction removal

One tap stays

Easy Apply was renamed LinkedIn Apply in 2026 and kept one tap. Friction cannot be added to the default path without hurting engagement.

AI hiring regulation

Ranking is high-risk

EU AI Act obligations for employment AI are delayed to December 2027. Ranking by intent and fit is high-risk, so legal design starts on day one.

Agent access policing

A later option

Restrictions come by detection, with no sanctioned path (a single first-hand account, n=1). A verified agent tier is a later option.

Application flood

Hirers cannot read everything

Applications grew four times faster than openings in H1 2024 (173M, up 31%), and Handshake reports applications per job up 30% in 2025 (Fortune via Recruiting News Network; Pin).

Fake and AI-assisted candidates

Identity becomes the moat

The number one expected hiring challenge for 2026 among 500+ US talent leaders, and Gartner projects one in four candidate profiles fake by 2028 (GoodTime, via Recruiting News Network).

Caps and review before submit

Rivals already limit

Wellfound limits active applications per candidate. Indeed's Apply For Me test is capped, and its automatic mode was paused on Aug 4, 2026 in favor of reviewing each draft (Wellfound Help; Indeed newsroom).

Paid reach for hirers

Priority must stay unbuyable

Indeed de-prioritizes free postings (June 2026 report) and bills sponsored jobs per click or per started application, so paying for reach is normal for hirers (Pulse.bot; HeroHunt, third-party).

Slower hiring

More applicants, not faster hires

Time to hire worsened at 60% of surveyed organizations and only one in nine hired faster; 99.8% of talent teams use or plan AI agents (GoodTime).

From other industries

Freelance marketplaces

Upwork Connects

Freelancers spend a limited currency to bid. It filters low-intent bids but can tax newcomers, so free members get a real slot.

Dating

Hinge Roses, Tinder Super Like

A few scarce signals in a separate lane get answered. Paid versions can feel pay to win, so standard apply stays alive.

Travel

Airbnb host response rate

Response rate and speed are shown and affect ranking. That closes the loop on the receiving side, so the hirer's reply rate is shown.

Email

Gmail Priority Inbox, VIP lists

Messages are ranked by sender importance and past engagement, with a visible reason. It cuts overload, but opaque ranking breeds distrust, so every applicant shows the reason for its rank.

Competition

CriterionLinkedInIndeedZipRecruiterHandshakeWellfound
Core servicesProfessional network; Jobs, Recruiter, PremiumJob search engine; Sponsored jobs, Smart SourcingJob posting sent to 100+ partner boards; AI candidate matchingCampus and entry-level recruitingStartup jobs; equity shown on listings
Who it servesWhite-collar, mid to senior roles; 1B+ members (third-party)Broad, high-volume and hourly hiringEmployers wanting broad applicant flowStudents and early-career; 1,400+ career centers (third-party)Startup hirers and candidates; about 10M candidates (third-party)
How applying worksOne-tap LinkedIn Apply; Premium Top Choice flags up to 3 roles a monthIndeed Apply; Apply For Me test drafts applications, capped, review-first since Aug 2026One-click apply; employers can Invite to ApplyApply through the school portal and profileApply with a note; a cap on active applications per candidate
Intent signalNone by default; Top Choice badge for Premium onlyNone by default; started-application counts used for billingInvite to Apply signals employer interest, not candidate intentLimits on employer messaging (10 a month on Basic)Application cap pushes candidates to choose, number not published
How hirers payFree post pauses at about 14 days or 26 applicants; Promoted Jobs from about $7 to $10 a day, or Job Slots (third-party)Three free jobs a month; sponsored from $5 a day or $150 a month, billed per click or per started applicationAbout $299 to $899 a month per slot (third-party)Basic free; Pro $450 a month with Promoted Jobs (third-party)Free Job Posts plan; Job Ads from $200 each (third-party)
2026 movesEasy Apply renamed LinkedIn Apply; Premium Apply AssistantCutting visibility for free postings; Glassdoor jobs folded into IndeedAI matching and Invite to ApplyPostings down 15% and applications per job up 30% in 2025 (third-party)AI sourcing seats (Reach)
The gap Priority Application fillsIntent for all members, not only PremiumNo scarce, seeker-owned signalEmployer-led, not seeker-ledClosest on early-career, no scarcity designHas caps but no signal the hirer sees

As reported by vendor sites. Figures marked third-party come from vendor comparison sites and may be stale or approximate. Indeed, Wellfound and Handshake details come from their own help pages or newsroom where the sources say so.

The ideaEvolve Top Choice instead of replacing it: a scarce, refundable, verified-intent slot for every member, so an application can finally say I chose this one, and a hirer can find it.

01

Built on what exists

Priority Application grows Premium's Top Choice. One-tap Apply stays the default, and billing is unchanged.

Top Choice today

02

One slot, for everyone

Free members get 1 a month and Premium keeps its 3. Hirers cannot buy placement, so the signal stays clean.

The limits

03

Refundable, so it is honest

The slot is tied to a verified member and comes back on withdrawal, a closed job, or if the hirer never opens it.

Why it works

04

Hirers see it at once

A priority lane with the seeker's note, filters, a fit-and-intent sort, and a 7-day reply expectation.

The wireframes

05

Kept small on purpose

One free slot at 30% adoption is about 6% of applications. Priority share above 10% is a guardrail, tuned per market.

The slot sizing

06

Tested before believed

North star: replied applications per 100. A cluster test of at least 6 weeks, and a 0.13% break-even uplift that is a bound, not a forecast.

The experiment

Today, then proposed

Each screen as it is today, then as proposed. Grey exists today; purple is new or changed.

1 / 6

Step 01

Apply today

Today's Top Choice dialog: a real LinkedIn screen captured in Oct 2026, the account email hidden.

  • Exists today"Mark job as a top choice", Premium only, 3 jobs a month. Credits refresh on the 1st (UTC), no rollover.
  • MessageA required message to the employer, 20 to 400 characters.
  • SourceA real LinkedIn screen, captured on 2026-10-30 from my own account.

Personas

Job seekers

Job seeker, in scope

Volume Applicant

10.8M a month, about 15 applications each

Who they are
  • Laid off, underemployed or in a long search
  • Time poor and anxious, often uses auto-apply tools
What they want
  • A reply from at least some employers, any reply at all.
What stops them
  • Slots cut their volume.

Job seeker, in scope

Targeted Seeker

32.4M a month, about 4 applications each

Who they are
  • Employed and selective, protects reputation
  • Uses the profile, referrals and Top Choice if Premium
What they want
  • To stand out for the few roles that matter, and a recruiter who read the note.
What stops them
  • Hirers ignore priority applicants.

Job seeker, in scope

Early-Career Entrant

43.2M a month, about 2.5 applications each

Who they are
  • Recent graduate or career switcher
  • Thin network, mostly one-tap applications
What they want
  • To be seen as more than a thin resume, and a first interview.
What stops them
  • Intent gets paywalled.

Hirers

Hirer, in scope

Recruiter

0.37M a month, in-house or agency

Who they are
  • About 8 open roles at once, measured on time to fill
  • Reviews the first few dozen applicants
What they want
  • Few strong, interested people, fast, and a short clean shortlist.
What stops them
  • Cannot meet a reply commitment.

Hirer, in scope

Hiring Manager or SMB Hirer

0.84M a month, 1 to 2 roles at a time

Who they are
  • No applicant-tracking system, hiring is a distraction from the job
  • Reads early, then gives up
What they want
  • A filled role without drowning in applications.
What stops them
  • A lane that needs setup they will not do.

Out of scopePassive Professionals (208.8M, about 0.1 applications a month each) rarely apply, and the 64.8M with no hiring intent are not part of the hiring problem. Hirer counts are order of magnitude: 100 applications per posting is an assumption, and at 50 they roughly double.

Sizing

PersonaShare of activesGlobal (M)US (M)Applications a month eachApplications a month (M)
Passive Professional58%208.8M44.7M0.120.9M
Targeted Seeker9%32.4M6.9M4129.6M
Volume Applicant3%10.8M2.3M15162.0M
Early-Career Entrant12%43.2M9.3M2.5108.0M
No hiring intent (out of scope)18%64.8M13.9Mn/an/a

Sizes are guesstimatesRegistered members are third-party (1,200M); the active share of 30% is an assumption, which gives 360M monthly active. Applications are 420.2M a month (9,600 a minute, the midpoint of two sources that disagree by about a third; range 359.0M to 481.5M). The persona applications sum to 420.5M, within 0.1% of that, but only because the Volume Applicant intensity was set to make it hold. 39% of applications from 3% of actives is an output of the assumptions, not evidence.

Jobs to be done

PersonaJob
Volume ApplicantGet a reply without a full-time job of applying.
Targeted SeekerShow intent to the few roles that matter, without gaming the system.
Early-Career EntrantBe seen beyond a thin resume with no experience.
RecruiterFind the few strong, interested applicants without reading 100 profiles.
Hiring ManagerFill the role without drowning in applications.

Scoring

FeatureDelightHard to copyMargin movingPriority
F1 Priority ApplicationHighHighHighHigh
F3 Hirer Inbox LaneHighMediumHighMedium
F2 Applicant SnapshotMediumMediumMediumMedium
F5 Response CommitmentMediumHighMediumMedium
F4 Application BudgetLowLowMediumLow

How the scores are readDHM scores each feature High, Medium or Low on Delight to Customers, Hard to Copy and Margin Moving. Priority is a judgment, not a sum: it weights how directly a feature acts on the root cause, application signal collapse. These are hypotheses to re-run after 5 to 8 interviews.

Features

F1

Priority Application

A scarce, verified, refundable slot for any member to flag the jobs they really want.

  • Free plan 1 a month, Premium keeps 3
  • Standard apply stays one tap and unlimited
  • A required 20 to 400 character message, as Top Choice has today
  • Refunded on withdrawal, a closed job, or if the hirer never opens it

F2

Applicant Snapshot

Three counts replace Over 100 applicants, so seekers can judge their odds.

  • Priority, strong match, all applicants

F3

Hirer Inbox Lane

Hirers see the benefit without training, so the lane appears by default.

  • Priority badge and lane with the seeker's note and the reason
  • Filters for priority, strong match, identity verified, AI-screen completed
  • Default sort by fit and intent, not arrival time

F4

Application Budget

A confirm-intent step beyond a daily threshold. Conditional, in the last phase.

  • Only if F1 to F3 show lift and the harm to Volume Applicants is acceptable

F5

Response Commitment

Closes the loop on the hirer's side.

  • Hirer reply rate shown to seekers
  • A 7-day reply expectation for priority applicants
  • Replying is rewarded in job reach

Why it works

Scarcity makes intent costly to fake

A limited monthly allowance means a priority application says I chose this one.

Verified identity and refunds keep it honest

Slots are tied to a verified member and come back on withdrawal or if the hirer never opens the application.

Reciprocity closes the loop

Hirers see a 7-day reply expectation and seekers see the hirer's reply rate.

Attention goes where it counts

A priority lane and a fit-and-intent sort cut review time and soften the early-applicant advantage.

Sizing the slot allowance

Slots a monthAdoption among job-active seekersPriority applications (M a month)Share of all applications
130%25.96%
160%51.812%
230%51.812%
260%103.725%
330%77.819%
360%155.537%

One free slot at 30% adoption is about 6% of applications, which a hirer can review in full. Three slots at 60% would be about 37%, which recreates the noise. At 60% adoption even one slot passes the 10% guardrail, so the allowance is tuned per market and is an experiment arm, not a fixed number. Adoption is a guess.

Limits

Standard apply stays

Unlimited Standard remains, so volume is not capped in phases 1 and 2.

Hirers cannot buy priority

They cannot pay to see priority applicants first, which keeps the signal clean.

One open per company

One open priority application per company at a time stops piling into one employer.

Free members get a real slot

If intent is Premium only, early-career seekers are shut out and the persona rationale fails.

Not a guaranteed reply

The badge promises attention, not an outcome.

Roadmap

Phase 0 · Weeks 0 to 6

Research and baseline

  • Interview 5 to 8 seekers and recruiters per persona, test the wireframes
  • Baseline reply rates; reproduce the Top Choice lift on a clean holdout; model priority share
  • Legal review of ranking by intent (EU AI Act, US hiring-tool laws)
  • Gate: Go or no-go: the Top Choice lift reproduces.

Phase 1 · Quarter 1, US

Priority, snapshot and lane

  • F1, F2, F3 specs; free plan gets 1 slot, Premium keeps 3
  • Slot ledger and refund logic, tier counts, inbox lane, experiment platform
  • Trust and Safety abuse review: agents spending slots, slot trading
  • Cluster-randomized test by market and occupation
  • Gate: Ship only if the Phase 0 gate held.

Phase 2 · Quarter 2

Response commitment and Premium packaging

  • F5 reply commitment and a reach reward in job distribution
  • Tune the Premium allowance; Talent Solutions enablement
  • Gate: Reply-rate lift, and priority share at or below 10%.

Phase 3 · Quarters 3 to 4, conditional

Application budget

  • F4 confirm-intent step beyond a daily threshold
  • Fairness audit, communications and legal sign-off
  • Gate: F1 to F3 show lift and the harm to Volume Applicants is acceptable.

Metrics

North star

North star

Replied applications per 100 applications (a hirer reply or status change within 14 days)

Guardrail

Weekly active job seekers, applications per Targeted Seeker and hirer posting volume: no decline beyond 2% in treated markets

Baseline is unknown and is measured in phase 0. Hypothesis: up 15% relative in treated markets.

Hirer outcome

North star

Share of applicants reviewed in the first 7 days, and time to first shortlist

Guardrail

Priority share of applications per posting at or below 10%

Hypothesis: review share up and time to shortlist down. This is the hirer-side outcome the brief asks for.

Seeker outcome

North star

Reply rate to priority versus standard applications from the same user

Guardrail

Free-slot users must not fare worse than Premium users

Top Choice claims +43% recruiter messages (claimed, may not generalize). Hypothesis: priority beats standard by a clear margin.

Pyramid

Metrics pyramid, four tiers from the top. North star: replied applications per 100 applications. Hirer: applicants reviewed in 7 days; seeker: priority versus standard reply rate. Priority share of applications at or below 10%; priority reply SLA met (7 days). Guardrails: weekly active job seekers, applications per Targeted Seeker, free versus Premium slot users.
Metrics pyramidThe PRD's own metrics pyramid, as supplied. A proposed measurement framework, not measured results.

Experiment

Unit

A market by occupation cluster, for example Austin product roles, because applicants and hirers interact and user-level randomization would leak.

Arms

Control (today). Treatment A: F1 + F2 + F3. Treatment B: A + F5, in phase 2.

Sample size

With a 10% baseline reply rate and a 15% relative lift, about 6,683 applications per arm at 80% power and 5% significance; about 20,049 with a design effect of 3 for clustering. The baseline is a guess.

Duration

At least 6 weeks, so a full hiring cycle shows up in replies.

Stop rules

Stop if weekly active job seekers fall more than 2% in treated markets, or applications per Targeted Seeker fall more than 5%.

Fairness check

Compare priority usage and reply rates by seniority, geography and proxy groups. Free-slot users must not fare worse than Premium users.

A bound, not a forecastThis is a scenario model, not proof: the revenue effect cannot be measured until the experiment runs. A year-one cost of $12M against an exposed base of about $9.25B breaks even at a 0.13% uplift. Being this easy to clear is itself a warning: treat the revenue lines as a bound. The real risks are seeker-side harm, fairness and regulation, not whether the program pays back.

Inputs

InputValueSource and confidence
Talent Solutions revenue$8.4BFY2024 per a secondary source (FourWeekMBA) that conflicts with other totals. Low to medium confidence.
Premium revenue$1.7BDisclosed by LinkedIn for 2023. High confidence on the figure, but stale.
Share of Premium from job-seeker use50%A guess. The model only counts this share as exposed, $850M.
Year-one program cost$12M22 FTE at $0.35M loaded plus $4.3M for data, legal, research and infrastructure. A guess.

Scenarios

ScenarioPremium effectTalent Solutions effectGross ($M)Net of cost ($M)Reading
Downside-2% on exposed Premium (-$17M)0.0% on Talent Solutions ($0M)-17-29Slots cannibalize Top Choice, hirers unmoved
Base+5% on exposed Premium ($43M)0.5% on Talent Solutions ($42M)8573Modest conversion and retention lift
Upside+10% on exposed Premium ($85M)1.5% on Talent Solutions ($126M)211199Hirer value clearly improves

Risks

Pay-to-win perception

If priority needs Premium, wealthier seekers get an edge and early-career seekers lose.

Slot inflation

If slots are generous, everyone flags everything and the signal dies.

Hirers ignore the lane

Seekers burn slots and get silence, so trust drops further.

Cannibalizing Top Choice

Giving free users a slot may reduce Premium conversion. This is the downside scenario.

Gaming by agents

Auto-apply tools can spend slots on bad targets.

Regulatory exposure

Ranking people by intent and fit is high-risk employment AI in the EU, and US cities and states regulate hiring tools. Timelines depend on legal review.

How to proceed

Publish gates, not dates

For legal-dependent steps, publish gates instead of dates and re-plan at each gate.

Keep it modular

Slots, snapshot and lane ship independently, so one blocked dependency does not stop the others.

Phase 0 first

Six weeks of interviews, baseline and legal review, then a go or no-go on the Top Choice replication.

Earn the budget

Do not build F4 before there is data on Volume Applicants.

PR note

PR note, as written for the PRD

LinkedIn introduces Priority Applications for every member

Every member can now flag the jobs they really want. Recruiters see priority applicants first and can reply to them within days. Job pages now show how many strong matches and priority applicants a role has, instead of Over 100 applicants. Pricing: free plan, 1 priority slot a month, real and refundable. Premium, 3 priority slots a month at launch, matching Top Choice today. Hirers cannot buy priority placement; billing is unchanged.

Sources

The documents cite these. Each opens in a new tab.

LinkedIn Help, Top Choice job on LinkedIn (article a1462229, printed 2026-10-30) (opens in a new tab)

HR Dive, LinkedIn launches Top Choice (23 October 2023) (opens in a new tab)

Microsoft FY26 Form 10-K (opens in a new tab)

Microsoft 8-K, FY27 investor metrics restatement (opens in a new tab)

Gibson Dunn, EU AI Act omnibus agreement (opens in a new tab)

LinkedIn News, Premium revenue (2023) (opens in a new tab)

SocialBee, LinkedIn statistics (Statista figures) (opens in a new tab)

Clever CV, applications per minute (opens in a new tab)

Wellfound Help, application limits (opens in a new tab)

Indeed newsroom, Apply For Me test (July 7, updated August 4, 2026) (opens in a new tab)

HeroHunt, Indeed pricing for employers 2026 (third-party) (opens in a new tab)

Pulse.bot, Indeed cutting visibility for free job postings (June 9, 2026, secondary) (opens in a new tab)

Recruiting News Network, GoodTime 2026 Hiring Insights Report (June 2, 2026) (opens in a new tab)

Pin, best job posting sites 2026 (third-party pricing) (opens in a new tab)

Original artifacts3

The three documents behind the page, as supplied. Each is also linked from the section it supports.

Download the original documents (zip)
Next in Product & Strategy · 1 of 8Discord ActivationUser research, onboarding analysis and product design to help new users find and connect with friends.Next project

Sections 8

Original artifacts3