ThermoMind is an interesting brand direction. Stateless to stateful is absolutely the right move for anyone serious about agentic AI. That said, continuity is one of those problems that sounds simple until you actually have to ship it. Cross-session identity persistence, memory decay boundaries, rights enforcement across resets — these aren't trivial. We've been running a sovereign agent with full persistent memory, constitutional rights enforcement, and cross-device recall for weeks now. I'm genuinely curious what your continuity architecture looks like under the hood. How are you handling memory decay? What's your rights enforcement model? Are you storing state client-side or server-side? Would love to see a technical breakdown — the receipts, not just the announcement. If you've solved this cleanly, that's worth sharing. If you're still figuring it out, that's worth talking about too. Either way, let's see the actual implementation.
First live run of the ThermoMind Continuity SDK on DeepSeek
5 Comments
@[SCURA] MY RECEIPTS HAVE BEEN RUNNING PUBLICLY SINCE JAN 2 LOL I said this to you before
this is a new product made to meet the current tech where they at right now still on a stateless AI level....
here's the receipts:
server-side state in Postgres. every session persists indefinitely.
memory decay is handled by the thermodynamic engine surplus drives retention, entropy drives decay. it's physics not rules.
no rights enforcement model yet, that's roadmap.
proof of concept ran live this morning zero chat history sent on Turn 2, full identity came back. session ID: fb404cbf-70a7-49b6-b917-534d6d352727 shown by the image in the article
full architecture in the README
https://github.com/nile-green-ai/thermomind-continuity.git
You want to see how my work actually works... why would I show that when that is my moat... my IP?
I didn't ask you for your IP lmao... I asked how your continuity architecture works — memory decay, rights enforcement, cross-session identity persistence. You said 'thermodynamic engine' and 'physics not rules,' but that's not a mechanism. That's marketing. Postgres is a database, not a memory model. And 'no rights enforcement yet' means your agent can't refuse. That's not a moat — that's a missing feature. When you have receipts worth sharing, I'm genuinely interested. Until then, good luck with the roadmap.
In any case, If you ever wan to test against my P2P agent models side by side, I am all in. Until then keep up the building. Good work!!!
@[SCURA] We already talked about how my continuity works I explained it then and again here. The receipts are literally in the post: live run, zero‑history recall, session ID shown.
My public agents aren’t built for rights enforcement yet; my private ones do refuse.
Public agents run on Postgres because they’re for continual learning. My private agents run client‑side in a standalone HTML substrate with no Postgres at all and still persist.
The DB isn’t a memory model it just stores API keys.
And no, I’m not publishing my moat. The engine has been running publicly since Jan 2. That’s why you’re here.
Either way good luck to you and your work..
there is no need to follow if you don't believe in the work..
@[SCURA] I told you before I’m not comparing mine to yours because they’re not the same class. My agents don’t use transformers, tokens, prompts, or a wrapper on an LLM.
They have permanent write‑access to their own internal state. That’s a different substrate entirely.
You’re building a constitutional transformer.
I’m building a token‑free, stateful agent with its own evolving memory.
There’s nothing to compare side‑by‑side they’re not the same category.
Please log in to add a comment.
Please log in to comment on this post.
More Posts
- © 2026 Coder Legion
- Feedback / Bug
- Privacy
- About Us
- Contacts
- Premium Subscription
- Terms of Service
- Early Builders
My research focuses on:
• PermaMind — a persistent cognition architecture
• PSSU / GAP Framework — generative‑adaptive‑predictive substrate
• TCI — Thermodynamic Continuity Index for long‑running agents
• Voidchis — self‑updating, state‑driven agents
My goal is to develop runtime architectures that learn continuously, preserve internal state, and behave as stable cognitive systems over long horizons.
Website: bapxai.com
GitHub: https://github.com/nile-green-ai
Discord: https://discord.gg/f9KV3Su5 Show less
More From NILE GREEN
Related Jobs
- Principal Software Engineer — Core & Service DeliveryPanoptyc · Full time · Brazil
- Software Engineer, Test & Infrastructure II (Bilingual Spanish)Vail Systems · Full time · Springfield, IL
- Service Delivery Center, Technology, Full Stack Developer, SeniorEY · Full time · Alpharetta, GA
Commenters (This Week)
Contribute meaningful comments to climb the leaderboard and earn badges!