AI Builders Domain - GG24 Sensemaking Report

[thedevanshmehta]

Problem & Impact

AI is everywhere, including in the onchain economy. Outcompeting the traditional world relies upon our ability to use LLMs in ways that are faster and better than the incumbents.

We thus propose a domain dedicated to projects launched on Ethereum that have plans for integrating AI into their workflow.

This will provide 3 benefits

1. Give a sense for how projects in the ecosystem are planning to use AI

2. Let us provide appropriate support for guiding upon its actual integration and deployment by teams

3. Accelerate AI usage on ethereum

Now that several foundational models have crossed the viability chasm for being good enough, we need to focus on their integration into existing products and services. If the decentralized world can better leverage the capabilities of LLMs as they exist today, we can make use of this short window of time to overthrow existing incumbents in the traditional world

Sensemaking Analysis

Pond recently ran a funding round where 76 high quality AI projects participated, with judges giving scores to each of them. We propose using their infra to accept applications from the ecosystem and a judging pool of experts to select winners from among them.

We anticipate this round as being especially useful for sourcing projects rather than funding. Almost every project in the ecosystem has a strategy for using AI in its operations or product. Being able to bring these ideas into one arena along with identification of the high potential teams in the ecosystem working on AI lets us provide more dedicated (financial and non-financial) support to make them a success.

Gitcoin’s Unique Role & Fundraising

We need to test better methods for integrating AI into capital allocation. For this domain, we are keen to keep a mandatory requirement that every applicant create an AI chatbot that has been fed sufficient context so that judges can simply talk to the AI for making decisions instead of having to read application forms (judges may optionally choose to see the context fed to the AI rather than solely use the conversational interface).

We will also require applicants to feed their AI with the latest context so that at any time evaluators can simply query the chatbot to measure and gauge progress on the milestones upon which more funding can be streamed.

This requirement has the nice feature of only letting teams with decent AI chops even being able to apply, while moving us beyond grant application forms into a conversational interface that can be auto updated by teams as they make progress towards their goals.

Success Measurement & Reflection

Positive outcomes comprise actual launch and integration of AI capabilities in applicant projects. We also see follow-on funding from investors and mentorship on LLM use by experts as an important component

Success is also measured in having a successful chatbot that can represent teams over time, marking an important step in moving beyond static grant applications for judges and evaluators.

Domain information

This domain is only open to teams that clearly list out their plans for integrating AI into their product or operations. They will also be required to submit a URL with an LLM that has information on their application, so that judges can learn about the team via a conversational interface instead of a static form.

We expect the LLM to be updated with context over the 6 month period so that evaluators can also interact with it to get updates on the project’s progress, based on which amounts are streamed to the teams over time upon milestone completion.

Devansh Mehta, Bill Shi and Davide Crapis will serve as domain experts, with additional judges being pulled in after a consensus of these 3 domain experts.

[owocki]

Draft Scorecard

2025/08/18 - Version 0.1.1

By Owocki

Prepared for Devansh Mehta re: “AI Builders Domain - GG24 Sensemaking Report”

(vibe-researched-and-written by an LLM using this prompt, iterated on, + edited for accuracy quality and legibility by owocki himself.)

Proposal Comprehension

TITLE
AI Builders Domain - GG24 Sensemaking Report

AUTHOR
Devansh Mehta (thedevanshmehta).

URL
https://gov.gitcoin.co/t/ai-builders-domain-gg24-sensemaking-report/23049

TLDR

Fund a domain focused on helping Ethereum projects actually integrate AI. Applications must include a working chatbot loaded with project context so judges can evaluate via conversation and later query ongoing milestone progress; use Pond’s infra and a panel of domain experts to select and support high-potential teams.

Proposers

Devansh Mehta
AI x Public Goods & Governance lead at the Ethereum Foundation; active on AI-for-funding experiments like Deep Funding.

Domain Experts

Devansh Mehta
AI x Public Goods & Governance, EF.

Bill Shi
Co-founder/CTO at Pond, a crypto-native AI startup that raised a $7.5M seed led by Archetype.

Davide Crapis
Research scientist affiliated with the Ethereum Foundation; cofounder of PIN AI; publishes on AI agents and crypto.

Problem

Many Ethereum projects want to use AI but lack structured support and fast evaluation pathways. The proposal argues that the window to outcompete incumbents is now, via practical LLM integration into existing products and operations.

Solution

Run an AI Builders domain using Pond’s infra and a judging pool. Require each applicant to submit a functioning AI chatbot loaded with their application context; judges converse with the bot instead of reading long forms. Require teams to keep the bot updated for six months so evaluators can query milestone progress and stream funds accordingly.

Risks

1. milestone verification: querying a bot for progress is useful but insufficient; must pair with verifiable artifacts, usage metrics, or code checkpoints; otherwise streamed funding could be misallocated.
2. infra dependency: reliance on Pond’s stack is efficient but creates single-vendor risk; have a backup application path and data export plan.
3. timeline: building and maintaining chatbots could slow initial participation unless templates are provided; risk to fast October delivery if setup is heavy

Outside Funding

No explicit co-funding is stated for the domain itself. Pond’s company is funded, but that is infra support rather than direct round co-funding as written.

Will this domain be impactful if the funding amount is small? AI builders literally have VCs chasing them right now. Do they need our $25k?

Why Gitcoin?

Gitcoin can convene a large cross-ecosystem pool of builders, run transparent program ops, and pressure-test novel capital allocation methods in public. This domain aligns with Gitcoin’s role as a design-space lab for funding mechanisms, and its network effects help with sourcing, mentorship, and distribution.

Owockis scorecard

#	Criterion	Score(0-2)	Notes
1	Problem Focus – Clearly frames a real problem, avoids solutionism	1	The problem is timely and important, but framed broadly. Would benefit from clearer sub-problems and target segments.
2	Credible, High-leverage, Evidence-Based Approach	1	Requiring bots is novel and potentially high-leverage; prior Pond round suggests signal, but more evidence on evaluation reliability would help.
3	Domain Expertise – Recognized experts involved	2	Strong team spanning EF research, program design, and AI infra.
4	Co-Funding – Beyond Gitcoin	0	None specified for the round. Encourage matching sponsors.
5	Fit-for-Purpose Capital Allocation Method	1	Expert judging plus conversational apps fits early AI integration, but needs guardrails for rigor, reproducibility, and fairness.
6	Execution Readiness – Can deliver meaningful results by October	2	Pond infra exists and experts are named, but mandatory bot work may slow applications unless templates and hosting are provided.
7	Other – vibe check and misc.	2	Positive, ambitious, aligned with ecosystem needs; moderate risk of selection theater without strong verification.

Score

Total Score: 10 / 14
Confidence in score: 70%

Feedback:

Major

• find cofunders.

Minor

• publish a clear judging rubric with weighted criteria and sample questions to standardize chatbot interviews.
• define streaming rules in detail: milestone types, required proofs, pause and clawback conditions, reviewer assignment.
• recruit at least two external co-funders or prize sponsors to increase signal and reduce single-source risk.
• add verification. pair chatbot evaluations with required artifacts: code links, demo videos, unit tests, analytics snapshots, and short writeups; include random audit of underlying sources to reduce overfitting and hallucination risk.
• ease the onramp. ship a reference chatbot template and hosted path so teams can meet the requirement in hours, not weeks; otherwise you select for prompt engineers over builders.

Steel man case for/against:

For

This program could quickly map and accelerate practical AI adoption across Ethereum. Forcing conversational, up-to-date project agents can reduce reviewer load, improve information access over time, and teach the ecosystem how to operationalize AI in funding and post-grant accountability. The experts and infra are already on deck.

Against

Chatbots can be gamed and may privilege style over substance. Without strong verification and guardrails, you risk rewarding the best demo rather than the strongest technical integration and user impact. The mandatory bot requirement may shrink the applicant pool and bias toward teams with more AI resources.

Rose/ Bud/Thorn

rose
Clear, differentiated mechanism that pushes the frontier on AI-native grantmaking and reduces reviewer burden with living, queryable applications.

thorn
No cofunding

bud
If you standardize templates, rubrics, and streaming proofs, this could become a reusable protocol for AI-assisted grants across multiple domains, not just AI Builders.

Feedback

Did I miss anything or get anything wrong? FF to let me know in the comments.

Research Notes

open questions: what exact chatbot format is required, who hosts it, how is context updated and versioned, and how do evaluators verify claims beyond the bot conversation.
future diligence: request a 1-pager on the evaluation rubric, a minimal reference bot template, a data-handling policy for sensitive context, and a list of provisional judges and potential co-funders.

[deltajuliet]

@thedevanshmehta you’re killing it during Sensemaking Szn.

Evaluated using my steward scorecard — reviewed and iterated manually for consistency, clarity, and alignment with GG24 criteria.

---

Submission Compliance

• All sections present: problem, sensemaking, solution, domain info
• Domain experts listed (Devansh, Bill Shi, Davide Crapis)
• Mechanism defined (Pond infra + mandatory chatbot interface + expert judging)
• No co-funding commitments yet
• Verdict: Compliant, but risky without co-funders and stronger verification guardrails

---

Scorecard Evaluation

Total Score: 10 / 16

Criteria	Score	Notes
Problem Clarity	1	Important and timely, but framed broadly (“AI everywhere”) vs. a crisp Ethereum wedge.
Sensemaking Approach	1	Builds on Pond’s round; novel chatbot requirement, limited evidence on evaluation reliability.
Gitcoin Fit	2	Gitcoin as a lab for new allocation designs fits well.
Fundraising Plan	0	No co-funding commitments; external anchors needed.
Capital Allocation Design	1	Conversational apps + milestone streaming is differentiated; needs strict rubrics + proofs.
Domain Expertise	2	Strong experts across EF, Pond, and research.
Clarity & Completeness	2	Clear mechanism and team; execution details (bot template, streaming proofs) need fleshing out.
Gitcoin Support Required	1	Ops help needed for bot hosting, verification, and milestone rules.

---

Feedback for Improvement

Where I align with Owocki

• Co-funding is the biggest missing piece.
• Verification must go beyond chatbots — require artifacts (code links, demos, analytics), and random audits.
• Publish rubrics + streaming rules early to avoid confusion.

Additional suggestions

• The mandatory bot is a double-edged sword: forces AI-native grantmaking but may shrink the applicant pool. Ship a hosted reference template to lower the barrier.
• Define what counts as “AI integration” (infra, governance tools, consumer dapps) to prevent dilution.
• Consider interoperability so this becomes a reusable protocol for AI-assisted grants across domains.

---

Conditional Support

Would support if:

• 2–3 co-funders or sponsors commit, and
• Evaluation rubrics + reference bot template are published, and
• Verification framework (artifacts + audits) accompanies chatbot evaluation.

This domain is creative and aligned with Gitcoin’s role as a design-space lab. With proper scaffolding, it can accelerate credible AI adoption across Ethereum; without it, it risks becoming a novelty demo round.