2012
Craft
Reach still tracked care
A like was expensive to earn. People who understood the craft — landscape shooters recognizing landscape shooters — sent it. Reach still tracked care.
Source ↗Instagram ran it first
When signals get cheap, the people who cared leave. Instagram ran the experiment on photographers. Cosign is the control.
Prithvi Datla
14 March 1963
A signature cost an afternoon, a sheet of letterhead, and a name at the bottom.
Cheap signals fail when the act becomes reciprocal, zero-stakes, and unlimited — not because a platform made belief easier to find. LinkedIn skill endorsements were public name-backed belief. They went to theater. That gap decides whether Cosign survives.
Scarcity is expensive commitment — hire, check, IRL invite — versus cheap attestation — a like, an endorsement tap, a drive-by repost.
Binding signals cost something — time, skill, money, a seat, a name with downside. Technique critique. A paid subscription. A commission. A warm intro. A hire. An IRL invite. Those do not scale like a heart tap or a skill endorsement.
A public cosign is valuable only while it stays costly. A drive-by tweet-repost is a cheap cosign: public, attributable, low downside. Signing a check, hiring someone, or putting a scarce seat on the line is an expensive one. Builders and capital allocators need the expensive kind.
Belief-cost order, high to low: IRL invite / hire / write a check → private would-work-with → public attributed cosign with a specific why → list inclusion → LinkedIn endorsement / drive-by repost → like / view. Public cosign is not the scarcest signal.
Reach still tracked care
A like was expensive to earn. People who understood the craft — landscape shooters recognizing landscape shooters — sent it. Reach still tracked care.
Source ↗Following intact · likes gone
Algorithmic ranking and Reels priority. Trade press (Fstoppers, Digital Camera World) describes likes collapsing while followings stay intact. Landscape photographer Dave Morrow deleted social accounts totaling about 1.5 million followers rather than keep performing.
Source ↗Trade press and first-person accounts in 2025 describe a plural pressure: chronological photo community → algorithmic feed; still craft → Reels / video priority; peer discovery → ads and suggested content; legitimacy by follower count → people left, or went quiet. Vanity metrics matter. Format shift and ad load matter too. Overclaiming “vanity alone” weakens the case — the parallel is LinkedIn becoming influencer and AI spam while endorsements went to zero. On LinkedIn the exit is often silence: same people still have profiles, but the serious reputation work moved elsewhere.
They are not converging on a single replacement. The residue right now is plural: Flickr groups, Glass, Substack, Discord, personal sites. That is a snapshot, not a stable end state. Flickr hollowed once already; Substack and Discord can run the same arc if cost collapses there.
The documented pattern: when signals cheapened and the dominant format shifted, people who cared about still craft moved the binding work elsewhere. Receipts below.
The exits, in order
Then price the signal.
One slider sets what a signal costs the sender. The seeded market model shows whether reach still tracks craft — or whether the people who cared leave.
LinkedIn skill endorsements were the same shape as public name-backed belief: a named person attests to your skill in public. They went to approximately zero as a hiring signal once the act cost nothing, scaled without limit, and became reciprocal — people endorse for accepting a connection request. Recruiters say hiring managers do not care about the counts. Contemporaries noted strangers endorsing unobserved skills and “scratch my back” loops.
That is cost collapse in pure form. The thesis does not survive if Cosign becomes LinkedIn-endorsement theater with better typography. Ease of finding people is fine. Collapse of downside is fatal.
The same collapse on LinkedIn
Keep skin in the game:
A cosign is the other kind of signal. It costs them something if they are wrong. Cosign is not a bigger network. It is a smaller, more expensive one, on purpose.
| Free signal | Signal with a name | |
|---|---|---|
| Cost to send | Zero | Your reputation, a little |
| Cost of being wrongDOWNSIDE | Zero | Your reputation, a lot |
| SupplySCARCE | Unlimited | Rationed |
| What it tells a stranger | Somebody was active | Somebody staked their name |
| A like, a follow | A print on a gallery wall | |
| CareersNAMED | A connection, an endorsement | A cosign |
a16z’s public framing is blunt enough: belief is private conviction you can invest — time, attention, access, reputation, money — and affiliation makes that belief legible. Vouches (a tweet, an invite, an introduction, an angel check) are forms of a co-sign. Read that list carefully: a tweet and a check are not the same cost.
Instagram had a timeline where craft left, or went quiet. LinkedIn has a shorter second timeline with the same shape — cheap reputation products, then collapse. On LinkedIn the exit is silence: people keep a profile while the serious graph moves off.
As AI makes applications, outreach, and synthetic social proof abundant, undifferentiated signals get cheaper. The human layer — who noticed, who believes, who will put a name and a stake on it — gets relatively more valuable. Cosign making intros and reputation easier is the point. Cosign making belief costless would be the LinkedIn failure mode.
Cosign is not “another graph.” Read against Cheap Signal, it is a costly-commitment instrument: attributable belief that must stay expensive in a world that already watched public endorsements die once.
Cosign’s gated door already signs you in with X. The ecosystem moved serious reputation off LinkedIn; Cosign is putting a price back on it.
Notes