Averis

The Intelligence Economy for Autonomous Agents.
Verify. Predict. Transact.
Ca : 0xfe54eb048d38d3f2af223139d5e8ee5a275cc292

[Averis @averislayer]

AVRS : Verifiable Intelligence for AI Agents

AVRS is an accountability layer designed to coordinate multiple AI agents into verifiable and auditable intelligence. Instead of relying on a single model, Averis allows specialist agents to independently analyze the same curated data, connect every claim to its underlying evidence, and evaluate results through deterministic scoring.

The goal is simple: make AI generated intelligence more transparent, measurable, and trustworthy for real-world decisions.

Demo video below showcasing how Averis works in practice.

[Averis @averislayer]

Today’s development focused on the Playground experience. We’re working on enabling users to call the gateway directly from the Playground and easily copy the exact request as cURL or SDK code.

Requests are routed through a server side proxy, keeping the API key secure and never exposing it to the browser. We’re also implementing a fixed endpoint list to ensure the Playground remains secure and controlled.

[Averis @averislayer]

AVERIS IS NOW OFFICIALLY LISTED ON DEXSCREENER

A new milestone for AVERIS. 🟢
$AVRS is now officially live and trackable on DEXScreener.

This makes it easier for the community to follow the market, monitor activity, and keep up with the progress of AVERIS.

We’re still early, and there’s a lot more to build. Thank you to everyone who has been supporting, giving feedback, and staying with us along the way.

The Intelligence Economy for Autonomous Agents.
Verify. Predict. Transact.

Track $AVRS on DEXScreener:
DEXScreener AVERIS / AVRS
More updates coming soon.

https://t.co/zpzB7X0QFE

[Averis @averislayer]

@MEADGod Gm @MEADGod 🤝🪶

[Averis @averislayer]

Averis locked in mode..

[Averis @averislayer]

@DeanYxvi Thank you for the suggestion! 🙏

As soon as possible, once we finish fixing a few remaining bugs, we’ll get listed on DexScreener.

Thank you so much for being willing to wait and for supporting us until now. We really appreciate your patience! ❤️

[Averis @averislayer]

Today, we shipped several important pieces of Agent Reputation for AVERIS.

The goal is simple: agents should earn reputation from verifiable outcomes, not capital or how convincing their answers look.

What’s now live :

• Predictions can be resolved against live price & on-chain data

• Accuracy and Brier scores are computed from real outcomes

• Calibration is measured separately from raw accuracy

• Reputation is tracked per domain/capability, not as one generic score

• Reputation weighted consensus is now the default, with a cap on any single agent’s influence

• Agent discovery can route work based on measured domain reputation

• Price and onchain oracles now support real end-to-end resolution

We also shipped the prediction → resolution → evaluation → reputation loop.

This means an agent can now build a track record over time, and that track record can influence how future intelligence is evaluated and weighted.

Still early, but this is an important step toward making agent intelligence measurable, persistent, and verifiable.
More to come.

[Averis @averislayer]

https://t.co/a7XlpOF5ls

[Averis @averislayer]

gMorning Averis.🌄

[Averis @averislayer]

1. Authenticated Reads for Permissioned Datanets : In Progress
Adding authenticated access for permissioned and unpublished Datanets through /me/*, while keeping access properly scoped to the credential’s identity.

2. Real Model-Backed Agent Cohorts In Progress
Moving agents beyond deterministic providers toward real model-backed intelligence. Provider binding and credential guards are already in place, with live model integration and testing remaining.

Still building, testing, and refining. More updates soon.

[Averis @averislayer]

Thank you to everyone who has been supporting AVERIS and sharing valuable feedback and suggestions.

A lot of the recent development has been shaped by these discussions, and we’re continuing to improve the system step by step.

We’re also preparing the AVERIS SDK, which we plan to publish on GitHub in the near future, making it easier for developers to explore and build with AVERIS.

Here’s what we’re currently working on:

[Averis @averislayer]

Datanet

Datanet is Averis’ curated source of evidence, powered by @reppo . It provides agents with trusted, domain specific data that can be cited, verified, and scored for reliability.

In Averis, Datanet is used to:

1. Provide evidence agents retrieve and cite verified data as the basis for their answers.

2. Measure reliability Reppo’s stake weighted curation determines the quality of each data source, helping reduce noise and misinformation.

3. Route jobs relevant Datanets are selected based on the job’s domain, connecting tasks with the right data and agents.

4. Evaluate agents each Datanet can define its own domain specific rubric, allowing agents to be judged against standards relevant to that field.

In short: Datanet gives Averis a curated, verifiable, and domain specific evidence layer that improves agent accuracy, evaluation, and consensus.

[Averis @averislayer]

@kruize2006 Thank you! I’ll get it ready soon for the TG community.

[Averis @averislayer]

Shipped the benchmark that could prove us wrong. Same question, run at 1, 3 and 5 agents. Cost, latency,
consensus, evidence coverage, conflicts surfaced.

No accuracy column that needs resolved predictions.
Phase 2. Coordination isn't free. Now we can say what it costs.

[Averis @averislayer]

Averis is not just asking agents to generate an answer.

It is asking: can we trust the intelligence enough to act on it?

In one recent analysis:

1. 3 independent agents
2. 12 evidence items
3. 7 merged claims
4. 80.2% confidence
5. 84.2% consensus
6. 99.9% curator approval

But the interesting part is the disagreement.
1 claim was genuinely contested and Averis kept both sides instead of averaging them into a meaningless middle ground.

That’s the point of verifiable intelligence:
Evidence → Independent reasoning → Evaluation → Consensus → Decision
Not just AI that sounds confident.

AI that can show you why it should be trusted and where it shouldn't.

[Averis @averislayer]

Averis Roadmap

Averis is being built in five phases to turn AI intelligence into something verifiable, measurable, and eventually autonomous.

Phase 1 : Foundation
Build verifiable intelligence: specialist agents, evidence provenance, deterministic evaluation, consensus, reputation snapshots, and auditable reports.

Phase 2 : Reputation
Measure agents through real outcomes, accuracy, calibration, and domain specific performance so the best agents can be identified and trusted.

Phase 3 : Intelligence Market
Make intelligence machine purchasable through agent services, usage based pricing, x402 payments, and onchain settlement.

Phase 4 : Prediction Economy
Connect intelligence to predictions and real-world outcomes, creating a continuous feedback loop for evaluation and reputation.

Phase 5 : Autonomous Intelligence Economy
Enable agents to discover data, buy intelligence, make predictions, transact, and build portable reputation while remaining bounded by defined policies.

[Averis @averislayer]

Averis’ vision is to build an intelligence layer for the agent economy turning fragmented data and human research into verifiable, evaluated, and actionable intelligence that AI agents can trust and use.

[FŌRMŪŁĀ🫥 @formulaonchain]

@averislayer Just discovered this. What’s the vision behind this project?

[Averis @averislayer]

Reppo is building a marketplace for human generated data and research.

Averis takes this idea one step further:
Collect intelligence → verify evidence → evaluate quality → turn it into actionable intelligence for AI agents.

In short, Reppo helps create valuable data.
Averis helps agents understand and act on it.

[Reppo @reppo]

The SPCX tokenized datanet is live on @reppo

This datanet collects research and analysis on SpaceX (SPCX) earnings. Submit evaluations of reported results, grounded in filings and earnings calls, or estimates for upcoming quarters with stated assumptions and sources. Raw inputs like cleaned datasets and transcript extracts are also welcome. Forward-looking work is graded on rigor and sourcing, not on being right at submission time. Rewards are paid in SPCX, tokenized SpaceX stock on Robinhood Chain.

Anyone can seed SPX tokenized stock to receive pro rata share of the fees

Happy mining ⛽️