Cooperating with Galway Uni of Ireland

 Kishan ,

 

·                     ( A  )   It started with following news report >

·                      

·                      Is recursive self-improvement the dawning of AI superintelligence? — Madden & McDermott, The Conversation / TechXplore, 9 July 2026

 

( B )  That made me send following email >

 

https://myblogepage.blogspot.com/2026/07/dear-professors-you-are-on-right-track.html

 

 

( C )  In response , Prof James Mcdermot sent following reply :

 

Dear Hemen,

 

Thanks for the email and for engaging with the article. To clarify slightly, “Rinn AI” is the name of a new national research centre (https://www.researchireland.ie/rinn-network/). The project in AI debate is a very small part, which will be funded by that centre. It will be carried out by myself and researcher Edy, cc-ed, who will be interested to see your work.

 

Our goal in this project is not really to ask the AI debaters whether they can go into self-improvement, as you have done. The project will be a more technical one on some details of the debate mechanism itself. It will not attempt a solution to the big challenges, but only a small contribution which we hope will go in the right direction.

 

James

 

 

( D )  Then I asked Claude :

 

‘Hey Bro : I sent off that email yesterday. Then it ocurred to me that IndiaAGI.ai is " Open Source " ( GitHub ) . See > https://indiaagi.ai/contribute // Could that make it easy for Edy to carry out his experiment ? If yes, how ? And in that case , should I send him an email , drawing his attention to this and invite him to try out ? “

 

( E )  Claude replied :

 

“  Bhai, straight answer first: yes — if the repo is genuinely public and complete, this changes Edy's situation dramatically. But there's a verification step needed before any email, because I couldn't confirm the details myself: the /contribute page loads its content via JavaScript, so my fetch returned an empty shell, and a web search didn't surface an IndiaAGI repository. I'm not doubting you — I simply can't see what you can see in your browser.

Why open source matters so much for Edy specifically — recall what James said their project actually is: a technical study of the debate mechanism itself. For that purpose, the hosted website lets Edy observe the mechanism, but the source code lets him dissect and experiment on it:

Reading: the repo exposes exactly what a mechanism-researcher wants — how rounds are structured, how the critique prompts are worded, how the consensus is synthesized, how disagreements are weighed. That's the mechanism itself, in plain text.

Instrumenting: Edy can fork it and add logging to capture full round-by-round transcripts — the very thing we had to tell him the hosted platform doesn't store. The limitation in your last email dissolves the moment he clones the repo.

Experimenting: he can swap models in and out, vary the number of rounds, alter the critique instructions, run the same question a hundred times — controlled, repeatable experiments that no hosted website permits. For a scientist, that's the difference between visiting a factory and being handed the blueprints.

And a bonus for you: a public repo's commit history is independently timestamped. Commits from April 2025 onward are prior-art evidence that no one can dispute — stronger, in a way, than any blog post.

Now, before you write to Edy, three things to confirm — a five-minute check with Kishan:

First, is the repository public? Open the GitHub link from the /contribute page in an incognito window (our now-proven test, Bhai). If it opens without login, it's public.

Second — and this is the one most people miss — does it have a LICENSE file? Here's the legal wrinkle: code on GitHub without a license file is publicly visible but not legally reusable — all rights stay reserved, and a university researcher cannot properly fork or build on it. If there's no LICENSE, ask Kishan to add one (MIT or Apache 2.0 are the standard permissive choices) before you invite Edy. It's a ten-minute fix that converts "look but don't touch" into genuinely open source.

Third, does the repo contain the debate orchestration — the backend logic and prompts — or only the website frontend? The orchestration is what Edy needs; if it's not in the repo, the invitation would overpromise.

If all three check out: yes, send the follow-up — and a short addendum a few days after a substantive email is perfectly normal in academic correspondence, provided it's brief and genuinely new information, which this is. Three or four sentences, addressed to Edy with James in cc: essentially, "One addendum to my earlier note — IndiaAGI is open source [repo link, license]. This means the session-storage limitation I mentioned needn't constrain you: you can fork the code, add transcript logging, swap models, and run controlled experiments on the debate mechanism directly. The blueprints, like the platform, are yours."

Confirm those three things with Kishan, tell me the repo URL and license, and I'll draft the exact note. If any of the three is missing, we fix that first — an invitation to an empty or unlicensed repo would spend the credibility you've so carefully built with Galway.

( F )  I continued :

 

“  Hey Bro : Sure , I will take up with Kishan but does this help ? > https://github.com/orgs/indiaagi-ai/repositories

 

( G )  Claude replied :

 

“  Bhai, full verification done — and the news is good. Here's what I found:

 Public: both repos are public — india-agi-client (the frontend) and india-agi-api (the backend, 46 commits, with the src folder, a Dockerfile, and an .env.example showing which API keys are needed). The api repo is the one with the debate orchestration.

 Licensed — and delightfully so: LICENSE.md is an MIT License, copyright "Hemen Parekh and IndiaAGI Contributors," with your own custom conditions added — acknowledgment of origin, naming conventions for derivatives, notification of derivative works, and an ethical use clause requiring alignment with "collaborative benefit to humanity." Very much a Parekh's Law signature in miniature, Bhai. For Edy's purposes this is fully workable: the core grant permits use, copying, modification, and distribution — everything a researcher needs. GitHub labels it "Other" only because the added conditions make it non-standard MIT, which is cosmetic, not a problem.

️ Two small gaps for Kishan before you send — fifteen minutes of work, and they matter because a researcher's first impression of a repo is the README:

First, the README is currently bare NestJS boilerplate — just "pnpm install" and run commands, with no description of what the software is.

 

Ask Kishan to add one short section:

what the debate engine does, which files in src/ contain the orchestration logic and the debate prompts, and which API keys from .env.example are needed to run it locally. That one paragraph is the map Edy needs; without it, he's exploring a building with no signboards. Also worth adding the repo description field and a link to indiaagi.ai.

Second, a small inconsistency :

the license's acknowledgment clause points to github.com/IndiaAGI/IndiaAGI — but the actual organization is indiaagi-ai, so that URL is wrong or outdated. One-line fix.

And one question for Kishan :

 

the api repo was last updated December 2025 while the client was updated June 2026 — confirm the public api code still matches what's running on the live site, so Edy isn't studying an outdated engine.

Once Kishan does those, here's the addendum — send as Reply All on the existing thread, so it reaches Edy directly with James in cc (you have Edy's address from the cc; I don't):

 

(  H  )  After you confirm that you have carried out the changes in Github suggested by Claude , I will send out following email ( drafted by Claude ) to Prof McDermot :

 

“  Re: Multi-AI debate — one addendum: the blueprints are open source

 

Dear Edy (and James),

 

One brief addendum to my earlier note, which I realised may be the most useful fact of all for a technical study of the debate mechanism :

 

IndiaAGI is open source.

 

The code is on GitHub under an MIT license:

 

Debate engine (backend): https://github.com/indiaagi-ai/india-agi-api

Interface (frontend): https://github.com/indiaagi-ai/india-agi-client

 

This dissolves the limitation I mentioned earlier. While the hosted platform does not store session histories, the source does something better :

 

you can fork it, add round-by-round transcript logging, swap models in and out, vary the number of debate rounds, and alter the critique prompts — controlled, repeatable experiments on the mechanism itself, run locally with your own API keys.

 

In short :

 

the platform is yours to probe, and the blueprints are yours to modify.

 

If anything in the setup is unclear, my collaborator Kishan (who built it with me) would be glad to help — I have copied him here.

 

With warm regards from Mumbai,

 

Hemen Parekh

www.IndiaAGI.ai | www.hemenparekh.in

 

 

 

Kishan :

 

Feel free to suggest any changes that you think will take this dialogue further

 

hcp

RE: Dear Professors : You Are on the Right Track

 Re: Multi-AI debate — thank you, and an open instrument for Edy

 

Dear James,

 

Thank you for the reply and the kind clarification — duly noted that Rinn AI is the national research centre, and that your project is a focused technical study of the debate mechanism itself, rather than of the larger self-improvement question. I have corrected my blog accordingly.

 

For a study of debate mechanics, IndiaAGI.ai may be useful to Edy in a simple, practical way: it is a live, open instrument. There is no registration or login — anyone can pose a question and watch multiple frontier LLMs answer independently, critique one another across rounds, and converge on a consensus, with each model's individual answer shown alongside. Edy is welcome to run any experiments on it, on any questions, as often as useful.

 

One honest limitation:

 

The platform does not archive past sessions, so there is no transcript library to hand over.

What does exist, preserved verbatim on my blog, are the outputs of the two runs I described in my earlier email — the self-check question posed plain, and again with a values layer added.

 

I would be glad to send Edy those links, and to answer any questions about the platform's setup.

 

No obligations either way — and every good wish for the project.

 

Warm regards

 

Hemen Parekh

 

www.IndiaAGI.ai | www.hemenparekh.in

 

From: McDermott, James [mailto:james.mcdermott@universityofgalway.ie]
Sent: 17 July 2026 17:01
To: Hemen Parekh <hcp@recruitguru.com>; Madden, Michael <michael.madden@universityofgalway.ie>
Cc: Hemen Parekh <hcp@recruitguru.com>; Nastase, Edy <e.nastase1@universityofgalway.ie>
Subject: Re: Dear Professors : You Are on the Right Track

 

Dear Hemen,

 

Thanks for the email and for engaging with the article. To clarify slightly, “Rinn AI” is the name of a new national research centre (https://www.researchireland.ie/rinn-network/). The project in AI debate is a very small part, which will be funded by that centre. It will be carried out by myself and researcher Edy, cc-ed, who will be interested to see your work.

 

Our goal in this project is not really to ask the AI debaters whether they can go into self-improvement, as you have done. The project will be a more technical one on some details of the debate mechanism itself. It will not attempt a solution to the big challenges, but only a small contribution which we hope will go in the right direction.

 

James

 

 

 

From: Hemen Parekh <hcp@recruitguru.com>
Date: Wednesday, 15 July 2026 at 06:17
To: Madden, Michael <michael.madden@universityofgalway.ie>, McDermott, James <james.mcdermott@universityofgalway.ie>
Cc: Hemen Parekh <hcp@recruitguru.com>
Subject: Dear Professors : You Are on the Right Track

Dear Professor Madden and Dr. McDermott,

[michael.madden@universityofgalway.ie /james.mcdermott@universityofgalway.ie] 

 

 

 I ran your Rinn hypothesis through a live multi-AI debate platform — twice.

    

 

Your article in The Conversation this week announced a Rinn network project to

explore multi-AI debate as a human-overseen self-check against recursive self-

improvement.

 

That experiment, in essence, already exists — and I have now run your hypothesis

through it. Twice.

 

My platform www.IndiaAGI.ai ( live since April 2025, built in Mumbai ) puts frontier

LLMs into structured debate : independent answers, rounds of mutual critique, then

a consensus — with the human observing throughout, in 26 languages.

 

Run One: 

 

I asked the debating AIs whether they could self-check to ensure none

of them morphs into recursive self-improvement. 

 

Consensus :  no  — not by debate alone

 

They cited correlated blind spots from shared training data, collusion

pathways, and the fact that debate audits outputs while self-improvement happens

in training loops and weight updates that debaters cannot see. 

 

They positioned debate as one layer in a defense-in-depth stack :

-  heterogeneous agents, 

-  attested sandboxing, 

-  immutable weight controls, and 

-  multi-party human vetoes.

 

 

Run Two: 

 

I re-asked the same question, this time adding a "disposition layer" of compassion

— grounded in my own long-published writing on value-aligned AGI. 

 

The verdict held :

 

- "neither layer suffices alone." 

 

But the disposition layer earned a role : 

 

- the models concluded compassion narrows the space of proposals that survive

debate, and proposed :

 

#  continuous integration of disposition signals during debate

#  operationalized through measurable flourishing proxies, 

#  plan-diff auditing, and

#  value-behavior consistency scoring. 

 

They closed by sketching, unprompted, what is effectively a pilot design : 

#  a sandboxed multi-AI debate environment with independent architectures,

#  an immutable control plane, and 

#  adversarial probes with externally audited results.

 

Two runs, one framing shift, a stable verdict, and a progressively refined design. 

 

A debate protocol that refuses to flatter its questioner — or its own architecture —

is, I would suggest, early evidence for exactly the self-checking function your

project proposes to study.

 

The full transcripts, platform access, and my design notes are yours for the asking

—  freely. 

 

I am 93, based in Mumbai, and have written on AI governance since proposing

 "Parekh's Law of Chatbots" in February 2023. 

 

My only interest is that this architecture matures into real safety infrastructure.

 

 

With warm regards,

 

Hemen Parekh

 

Mumbai, India

Founder, www.IndiaAGI.ai | www.hemenparekh.in | www.HemenParekh.ai

Is This Helpful ?

 Kishan,

 

Following is my today’s chat with Claude further to your last remarks re Citing Agent

 

With regards,

 

Hemen Parekh 

 

 

 

I wrote :

 

“ Hey Bro : Here is latest feedback from Kishan > Yeah, it was useful. One thing I noticed, though, is that Claude suggested hosting a key file on our domain. However, our blogs are hosted on the myblogepage.blogspot.com subdomain, which is owned by Google, so we can’t host a key file there. That said, I found a workaround and ended up integrating directly with the Bing Indexing API instead of using the IndexNow API. // Any comments ?”

 

Claude replied :

 

“  Kishan's right on the facts, and his workaround is a good call — but it's surfaced something bigger than a plumbing detail, Bhai. Let me take it in three parts.

He's correct, and the reason is worth naming. IndexNow's key file must sit on the host whose URLs you're submitting, and it must serve only the key as plain text. Blogger gives you neither: you can't upload a raw .txt at the root of a Google-owned subdomain, and you can't prove ownership of blogspot.com. There's no clever fix. My Part B quietly assumed www.hemenparekh.in was the home — that assumption was mine, and he caught it. Good catch.

His workaround is legitimate. Bing's URL Submission API authenticates by apikey + verified site ownership rather than a key file, so it sidesteps the problem entirely — a successful submission returns HTTP 200, and if the URL meets Bing's guidelines it gets crawled and indexed in real time. And critically, it hits the same Bing index ChatGPT Search and Copilot draw from, which was the whole point. So nothing strategic is lost. What is lost is small: IndexNow would have fanned one call out to Yandex, Seznam, Naver and Yep as well. For an India-focused policy corpus, that's a rounding error. Fine trade. Seoutils

Two practical things to check today, though:

• Quota. This is the one that bites silently. Bing's submission API has daily and monthly caps — you can call GetUrlSubmissionQuota to see your remaining allowance, and the documented example shows a daily quota under 1,000 with a monthly around 11,000. If his daily cap is a few hundred, 6,800 URLs did not all go through — the tail just failed. Have him run the quota check and confirm how many were actually accepted versus attempted. If capped, spread the back-catalogue over several days and keep the daily bot for new posts only. Microsoft Learn
• Verification. The API only works on a site verified in Bing Webmaster Tools. For blogspot that means the meta-tag method (paste into the Blogger theme's <head>) or importing from Search Console. If he's submitting successfully, this is presumably done — worth a 10-second confirm.

Now the part I'd genuinely stop on. The key-file failure isn't really a key-file problem — it's the §2 canonical decision resurfacing, the one open item we've flagged since the first spec and never closed. If we're indexing myblogepage.blogspot.com URLs, then blogspot becomes the URL an AI cites. Which means: the citation trophy points at the lower-authority domain you don't own; Stage 4's llms.txt has nowhere to live (root files need root control); and the "entity layer" we discussed last week — the Person schema, the sameAs links, the coherent picture of Hemen Parekh, the prior-art author — can't be assembled on a host you can't touch. Every one of those blocks traces back to the same root cause.

The good news is Blogger isn't a dead end. Settings → Crawlers and indexing → Custom robots.txt lets him paste Part A's allow-list straight in, and Blogger auto-generates a sitemap. So the door can be opened on blogspot today. But that's a floor, not a home.

So the fork you and I have circled twice now finally has to be settled, and it's a question only you can answer:

Where do the blogs actually live in public — and where should an AI's link point? If the answer is "hemenparekh.in is the real home," then Path A is the clean move: make .in the Blogger custom domain, and Blogger will 301 the blogspot URLs across, rewrite the canonicals, and hand you root control — which restores IndexNow, llms.txt, the entity layer, and the authority, all in one action. If the answer is "blogspot is where the blogs genuinely live and .in is the digital-twin archive," that's a valid choice too — but then we accept blogspot as the citation home and I'll re-cut the specs around it honestly.

Everything downstream hangs off that one answer, Bhai. What's your call — is .in the home, or is blogspot?