@DanFessler@milkgangie I know nothing about this so let me see if I have this right
The op looks like actual 3d models that are just being “blown up” to show the “sprites” but they don’t look like they’re 2d sprites to me, it just looks like sections of the 3d model split up
Do I have that right?
Oracle has secured all the funding we need to fulfill our contract with OpenAI. With our final capital raise earlier this month, we became the first hyperscaler fully funded for this investment cycle. The financing we outlined at Oracle Investor Day, including vendor financing, is in place. Any further private discussions with vendors are just additional options. Our ability to deliver on our OpenAI contract does not depend on those discussions. Any media reports that will come out and say the contrary are incorrect and wrong.
Oracle and OpenAI are making rapid progress on our infrastructure partnership. Our Abilene campus, where Astra-6 was trained, is nearly complete. Shackelford, our next gigawatt-scale campus, is fully permitted. The handover of NVIDIA Vera Rubin racks in its data halls is on track for the coming weeks.
We continue to expect positive free cash flow in the second half of FY2028. We are targeting more than $30 billion in free cash flow in FY2029 and expect to self-fund future growth beginning that year.
@Alex_Kaasten I experienced this a little bit but it’s really really noticeable with gpt 6.2. Not much of a problem with other models or previous gpt. No idea why. /goal helps
@lex_node@omarchy If it’s the new ethereum how do I invest
Seems odd to claim Apple is the aol of the agent era considering their processors are particularly well-suited for ai purposes, to such an extent that omarchy is prioritizing Apple silicon compatibility
@GraemeMoore@AgentChud This is probably a dumb question but how do you know if something is quantum safe without actually having access to a quantum computer + the ability to crack encryption with it
I built a firewall that stops your AI agent from ever seeing your secrets, and it's live for @NousResearch Hermes Agent.
Here's the uncomfortable truth about AI agents: the moment you paste a seed phrase, a private key, an API key or a 2FA code into one, that secret is sent straight to the model provider, and it stays in your session history. You've handed your most sensitive data to a third party, often without thinking about it.
I wanted to fix that without making the agent useless. So I built Seedguard.
⚙️ What it does
Seedguard is a secret firewall for Hermes Agent. The instant a secret appears in anything heading to the model, whether a chat message, a file you cat, or the agent's own context, Seedguard swaps it for a placeholder like [[sg:SEED_3fa9c1]] before the request ever leaves your machine. The real value is held only in memory. It is never written to disk, never sent to the model.
In the demo, I hand the agent a seed phrase and ask if it can see the words. It answers, on its own: "I only see the placeholder, not any actual words. Seedguard intercepted and redacted it before it reached me." The model itself tells you it's blind to your secret. That's the point.
🔑 But your tools still need the real thing
A redactor that blocks everything is useless. So Seedguard has a second half: when a tool genuinely needs a secret, you arm it with /seedguard arm <id> 5, for one specific tool, for a few minutes. Only then is the real value swapped back in, only inside that tool's arguments. In the demo, after arming, the terminal runs with the real 66-character private key and reports its length, while the model still only ever saw the placeholder. The secret reached the shell. It never reached the model.
🧪 What it catches
Crypto-grade, verifiable detectors: BIP39 seed phrases (checksum-verified, not just "12 words"), EVM and raw hex keys, Bitcoin WIF and xprv, Solana and NEAR keys, Sui suiprivkey1…, Ethereum keystore JSON, PEM keys, provider API keys (OpenAI, Anthropic, AWS, GitHub, Stripe…), .env secrets and one-time 2FA codes. Transaction hashes and public addresses are left alone, so it doesn't break your crypto workflow.
🛡️ How it's built
Public Hermes surfaces only, with no patching of the core. Fail-closed: if the redactor ever errors, it sends a content-free stub instead of your data, so a bug can never leak a secret. Zero runtime dependencies, pure Python standard library. 129 tests, ~93% core coverage, validated against the latest Hermes (0.23).
🙏 And it's honest about its limits
Hermes' auxiliary side-model calls (titles, compression) don't run plugin middleware yet, so a CLI-typed secret could reach them. Instead of hiding that, Seedguard ships /seedguard doctor, which detects it and prints the exact config fixes.
This is my entry for the Hermes Agent Plugin Hackathon.
Open source, MIT:
📦 https://t.co/mvOe932lV4
🧩 Catalog PR: https://t.co/177LsG8Idt
Your agent shouldn't know your seed phrase. Now it doesn't.
@AgentChud i've only tried sol so far, but it seems like i have to push it to actually do things. Like it can't derive from context that I want it to go ahead and do something, and I have to keep reminding it. It's good though and very cheap.