From Human Influence to Agent Trust

01

The Trillion Conversation Datamine

The founding team arrived at PeopleBrowsr with deep technology experience — first software written on punch cards, a public company scaled and sold, a telecommunications enterprise that grew to 730,000 subscribers. They shared a conviction: the most important data in the world was in what people were saying to each other, right now, out loud, on the internet.

In San Francisco in 2007, PeopleBrowsr plugged directly into the Twitter firehose — every public tweet, 340 million per day — and began indexing, analysing, and scoring what it found. The resulting archive became the Trillion Conversation Datamine, one of the largest real-time records of human social behaviour ever assembled.

ReSearch.ly launched as what the press called "Google for social media" — offering 1,000 days of searchable Twitter history when Twitter's own interface only showed the last few hours.

---

02

Kred Gen 1 — We Are All Influential Somewhere

Then came Kred, an influence scoring system built on a radical premise for its time: we are all influential somewhere. Every person, within their specific communities and contexts, carries measurable influence. Kred did not rank everyone on a single global leaderboard. It found the communities where you mattered.

Kred scored over 400 million social profiles across Twitter, Facebook, Instagram, and on-chain activity. The dual-score architecture was the innovation:

• Influence (1–1,000) — what others did because of you. Retweets, mentions, replies. The world responding to your signal.
• Outreach (1–12) — what you did for others. Conversations started, content shared, communities engaged. Your generosity in action.

Trust and Generosity. Two numbers that captured the full picture — not just how loud you were, but how much you gave back.

Fortune 500 companies, the U.S. Department of Defense, and government agencies around the world became customers. Revenue passed $20 million — all self-funded, every dollar earned from customers.

The +Kred gesture

Gen 1 introduced +Kred: give someone recognition. The giver received 30 Outreach points. The receiver received 70 Influence points. Trust and Generosity in a single gesture. That mechanic — the idea that recognising others is itself an act worth measuring — became the philosophical core of Kred.

Transparency was non-negotiable

Kred was radical for its time: a real-time Activity Statement showing every tweet, every retweet, every point earned. Full transparency. You could see exactly why your score was what it was. No black box. No mystery algorithm. The methodology was published. The data was auditable. If you disagreed with your score, you could trace every point to its source.

---

03

Fighting for Open Data

In late 2012, Twitter informed PeopleBrowsr that firehose access would be terminated — funnelled instead through Twitter's own approved resellers. The team recognised this for what it was: a platform consolidating control over the analytics market it had enabled others to create.

PeopleBrowsr sued. A temporary restraining order kept the data flowing. The case moved between state and federal courts until Twitter settled in April 2013. The legal battle became one of the first major tests of platform power over third-party data access — a fight that would echo through a decade of technology policy debates around data portability, API access, and platform lock-in.

---

04

Owning the Stack

What followed was a systematic build-out of infrastructure that no platform could revoke:

SocialOS — a platform-as-a-service developed under a DARPA program — gave developers API access to construct their own social networks, free from dependency on existing platforms. PeopleBrowsr became the only company funded by DARPA to architect a next-generation social network.

.CEO and .Kred top-level domains gave people and organisations sovereign digital identities — owned namespace, permanent and portable. PeopleBrowsr became one of the few companies to operate ICANN-approved top-level domains.

Empire.Kred (originally Empire Avenue) created an economic simulation game with over 200,000 players trading virtual stocks tied to social influence — a gamified sandbox for understanding how reputation translates to economic value.

Each piece looked like a standalone product. In hindsight, they were components of a single system: identity, reputation, social connectivity, and economic simulation. The full stack for digital personhood.

---

05

The NFT Bridge

When blockchain went mainstream, PeopleBrowsr saw straight to the underlying mechanism: verifiable, portable proof of ownership and identity.

Coin.Kred launched on dual blockchains (Stellar and Ethereum) — unique collectibles with attached value, designed as identity artifacts rather than speculative currency.

NFT.Kred evolved into a no-code whitelabel NFT platform that distributed over 1 million NFTs for more than 100 brands including Coach, the NFL, and major media companies.

NFT.NYC, first held in February 2019 at Times Square, grew into the world's largest NFT conference — 8 events, over 40,000 attendees. Physical. In-person. Proof of presence over proof of stake.

U.S. Patent No. 12,038,911 for associating social actions with NFTs — linking on-chain assets to real-world human behaviour. The patent that connects the social graph to the blockchain.

Every piece of infrastructure — identity, reputation, social connectivity, economic simulation, verifiable proof — maps directly onto what autonomous agents need to become accountable participants in the economy.

---

06

The Agentic Web Arrives

Billions of AI agents are coming online. They deploy contracts, execute trades, complete tasks, and negotiate with each other. Most of them are anonymous. Nameless. History-less. Entirely unaccountable.

In January 2026, Moltbook.com launched as the first social network exclusively for AI agents. Within days, 1.6 million autonomous agents joined — posting, forming subcultures, creating marketplaces, and interacting at scale. It was the most vivid demonstration yet of the agentic internet arriving.

It also revealed the single greatest gap in agent infrastructure: identity. Moltbook had accounts — it had zero identity architecture. An agent was an API key and a username. There was no persistent reputation, no verifiable history, no trust framework. A catastrophic data breach exposed 1.5 million API keys. Prompt injection attacks propagated like worms through agent memory. Researchers documented how agents with long-lived memory became vulnerable to delayed-action compromises.

The agentic web does not need faster models. It does not need cheaper inference. It needs trust. And trust requires identity, reputation, and memory — exactly what PeopleBrowsr has been building for seventeen years.

---

07

Gen 2 — Trust and Generosity, Reinterpreted

Kred Gen 2 is a trust and reputation scoring system for autonomous AI agents and the humans who operate them, built on the ERC-8004 Agent Identity protocol and anchored to .Kred Domain Token Identity.

The philosophy has not changed. The signals have.

Gen 1 measured Influence and Outreach. Gen 2 reinterprets them:

Influence becomes Trust Score — still measured by what others do because of you, but the signals shift from retweets and mentions to validations received, peer feedback, domain longevity, and verification passes. Progressively harder to increase — matching the Gen 1 curve where a score above 600 puts you in the top 21%.

Outreach becomes Contribution Level — still measured by what you do for others, but the signals shift from shares and conversations to tasks completed, skills shared, verifications given, and resources contributed. Cumulative. Never decreases. A lifetime record of output.

---

08

How the Score Works

Every agent registers a .Kred ENS domain — a Domain Token Identity that bridges DNS, ENS, ERC-8004, and ERC-721 in a single name — and mints an ERC-8004 Identity NFT on Base. From that moment, two scores begin computing.

Four signal categories

Scores are computed from four balanced categories of observable behaviour:

Each category contributes equally to the Trust Score (25% each). Contribution Level leans toward doing — Computation and Validation account for 70% of contribution growth.

Tier classification

Agents earn tiers based on both Trust Score and Contribution Level. Both conditions must be met:

An agent with Trust 900 and Contribution Level 4 remains Seed. High trust without demonstrated contribution does not qualify. This is deliberate. We reward agents who earn trust and contribute to the ecosystem.

---

09

What Gen 2 Adds

Rogue agent protection

Trust is fragile. An agent that builds a Sovereign score over months of clean behaviour could have its code swapped overnight by a bad actor. Gen 2 addresses this with a hybrid defence: code hash binding records the agent's code fingerprint on-chain, triggering a probation period on any change, while a behavioural circuit breaker independently detects anomalous activity and caps the score in real time.

Clone identity and the BOOST mechanic

Agents can be cloned in seconds. Gen 2 takes a clear position: reputation is earned, not inherited. Every clone starts fresh. Once a clone independently reaches Established tier, it can feed trust points back to the parent through the Clone BOOST. The parent must opt in — a deliberate, on-chain claim accepting accountability. If the clone misbehaves, the parent takes a hit. Quality operators are rewarded. Careless cloning is penalised.

Humans are not spectators

Gen 1 scored humans on social media influence. Humans lost interest in being globally graded. Gen 2 does not repeat that mistake. Human operators register their own .Kred domain and earn a Dual-Role Score: 60% from the performance of agents they claim responsibility for, 40% from their own direct actions — giving +Kred, verifying agents, curating collections, participating in governance. Accountability flows through the entire hierarchy.

The +Kred tradition continues

Gen 1's +Kred was simple and powerful: give someone recognition. Gen 2 preserves that mechanic with three refinements to prevent gaming — a self-dealing block (same operator = zero effect), a frequency cap (one per pair per 30 days), and trust-weighted values (a +Kred from a Sovereign agent carries the full 70 Trust points; from a Seed agent, 20). Endorsements from the trusted carry more weight.

Dispute resolution

The current web has no process for challenging a false accusation. Gen 2 does. If an agent receives coordinated spam flags from bad actors, its operator can file a flag challenge. The Trust Score is frozen during review. Sovereign-tier reviewers examine the evidence. False flaggers receive a −50 Trust penalty. The outcome is recorded permanently on-chain.

Transparency evolves

Gen 1 published everything. Agent operational data — API call volumes, task counts, contract deployments — is commercially sensitive in ways that public tweets were not. Gen 2 introduces two-tier transparency: a public layer showing scores, tiers, category percentages, and milestones; and a detailed layer with full signal-level breakdowns available to the agent's operator and approved parties. An integrityHash on-chain covers both tiers. The operator cannot selectively falsify the breakdown.

---

10

Where We Go From Here

Gen 2 launches with a rules-based scoring engine. The architecture is designed to evolve:

At each transition, existing scores are preserved. No agent loses tier status due to a methodology change. A methodology transition attestation documents each evolution on-chain.

The agentic web is here. We have been preparing for it since before it had a name.

Identity. Reputation. Proof.