A polished and functional GPT5, Claude4, Gemini (and more) AI interface and API that runs on a pay-per-query model via Lightning payments.
Public Key
npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h Profile Code
nprofile1qqsdy27dk8f9qk7qvrm94pkdtus9xtk970jpcp4w48k6cw0khfm06mspz3mhxue69uhhyetvv9ujuerpd46hxtnfduqs6amnwvaz7tmwdaejumr0ds4e7t45
Show more details
Published at
2025-08-08T15:00:19Z Event JSON
{
"id": "e25c46f4a98d870887428c58c3afb1dbcc3d7ebd2e239056f15abba10f2d2de8" ,
"pubkey": "d22bcdb1d2505bc060f65a86cd5f20532ec5f3e41c06aea9edac39f6ba76fd6e" ,
"created_at": 1754665219 ,
"kind": 0 ,
"tags": [],
"content": "{\"name\":\"PPQdotAI\",\"about\":\"A polished and functional GPT5, Claude4, Gemini (and more) AI interface and API that runs on a pay-per-query model via Lightning payments.\",\"lud16\":\"[email protected] \",\"nip05\":\"[email protected] \",\"picture\":\"https://m.primal.net/JNEu.png\",\"displayName\":\"PayPerQ\",\"display_name\":\"PayPerQ\",\"website\":\"PPQ.AI\",\"banner\":\"https://m.primal.net/JcwM.png\"}" ,
"sig": "fdda1a8b06caf2c1185634badc692c9fc3c453f54b70fe5094e02dcaa024eecbc86894a274b9ea8bbd635280479615c92530a8964774ef155e4ee059ffd94403"
}
Last Notes npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI Many customers asking how to red-team their own codebase. We just added the @abliteration_ai models (GLM 5.3 and DeepSeek V4.1 Flash, abliterated) Refusals are removed from the weights, so they work through your security issues instead of declining! https://x.com/PPQdotAI/status/2115110358184534247?s=20 npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI We've got internal access to this model and its FIREEEEEE. We're told we'll be able to roll this out to users in a matter of a week or so. 🤞🤞🤞 https://blossom.primal.net/ecdb6f7bca0b44d67ab91d825487ca49ed599fc1f2fb4a7e3a88d708b7f49804.png npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI New on PayPerQ: PII Redaction. Names, emails, and IDs are stripped from your prompt before it reaches the big AI providers, then restored in the answer. All inside a secure enclave neither PPQ nor our PII provider TinfoilAI can see into. Live free! https://ppq.ai/blog/pii-redaction npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI Does your harness ever encounter a javascript rendering or bot detection issue when using standard curl or webfetch attempts? If so, you can tell your harness to use firecrawl in our data enrichment endpoints to get the scraping done! Example harness instructions: **Blocked web pages — escalation order:** when WebFetch or `curl` returns a bot-detection or "checking your browser" page instead of content, try in this order, stopping at the first that works: 1. Firecrawl through our own endpoint (`POST /v1/data/api/firecrawl/scrape`, body `{"url": …, "formats": ["markdown"]}`, ~$0.015/call). It renders JavaScript and handles standard Cloudflare-style challenges, so it is the right second attempt for any JS-rendered or bot-gated page. 2. The site's own alternative front door: a `noscript.` subdomain or link in the challenge page, an onion/I2P mirror, an RSS/sitemap/`llms.txt`, or its public API. 3. A cached copy: `curl` the Wayback Machine (`https://web.archive.org/web/2026id_/<url>`, pass `--compressed`); WebFetch cannot reach http://archive.org. Do not loop on the same fetcher with longer waits — one retry is enough evidence that it will not pass. Head to our data enrichment page to learn more! https://ppq.ai/api-docs#data-enrichment npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI Yea. The "free" was being abused. A few users started trying to burn down thousands of dollars a day creating hundreds of videos each. npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI Google's image and video models are 95% off on PPQ for the next couple of weeks. Nano Banana 2.1, Nano Banana Pro, Veo 3.1, Veo 3.1 Fast and Gemini Omni 1.1 Flash. A Veo clip for about 2 cents, an image for a fraction of one. Pay-per-use, no subscription! PS: Yes, it's real! https://blossom.primal.net/e7fbd4ec75b7a6911453563dec90e976fb6d308bc75978e2e7bcf2dc7984ff71.mp4 npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI Thank you for the feedback. We just pushed a fix on this. Can you please tell us if it is to your liking? npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI Gemini 4 Argon has arrived at PPQ. and we're slashing the price heavily: 75% off of the list price (50% off the price that even Openrouter charges) $1 input and $5 output per 1M tokens While supplies last! Go nuts! https://blossom.primal.net/4774cf70cd4f005d7fecc6a81e2320f0e40e1fefdafd8df4b0f33c9fe0ec8b3f.png npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI 🫡
#nevent1q…my28 npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI Hello, I believe you may be one of the affected accounts to a security incident we had this weekend. A few dozen users were affected and we are reimbursing them. Very sorry for the inconvenience. https://ppq.ai/blog/referral-link-security-incident npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI Thank you. Appreciate the warm comments. Additionally, we just launched our Enclave Router today which is a huge leap for user privacy: https://primal.net/e/nevent1qqsq22x07zz4crar0e4s9299y4v2zkr74ejmua3chfshfsswrghylrg8lnwrs npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI Thank you for the feedback npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI 14/ We believe the entire routing industry will eventually move towards enclaves. We are happy that we are among the first, but it is possible that security researchers may find some issues or points of improvement for our particular implementation. We welcome any and all feedback. Please reach out to us! npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI 13/ PPQ exists to put the best AI in the hands of anyone, anywhere, safely and privately. No account, no subscription, no surveillance. The Enclave Router is the biggest step we've taken toward that goal yet. We will continue building in this direction forever. npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI 12/ Full writeup, including exactly what we can and can't see: https://ppq.ai/blog/introducing-the-enclave-router Nothing to turn on. It's the default for every user, on the web app and the API, starting today. If you want to PROVE the full verification for yourself, use the verifier above! https://ppq.ai/blog/introducing-the-enclave-router npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI 11/ Honest limits: this covers chat today. Image, video and audio still hit our backend directly; bringing them in is next. Frontier providers like Anthropic still see your prompt, because that's where the model runs. For our E2EE models on tinfoil.sh, not even the provider does. npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI 10/ Add the rest of PPQ: no accounts, email optional, pay with Bitcoin, Lightning, Monero or stablecoins. The provider that can read your prompt doesn't know who you are. PPQ, which knows how you paid, can't read your prompt. Content and identity never sit in the same place. npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI 9/ We also shipped the PPQ Privacy Verifier: a small open-source proxy you run locally. Point any OpenAI- or Anthropic-compatible client at it. It fetches the enclave's attestation, checks it against the published build, and refuses to send on a mismatch. https://github.com/PayPerQ/ppq-privacy-verifier npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI 8/ Don't take our word for it. The enclave code is open source and reproducibly built, so anyone can check that production runs the code we published: https://github.com/PayPerQ/ppq-enclave-proxy npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI 7/ Model fraud is ruled out the same way. The enclave's code binds each model family to its provider. A Claude request can go to Anthropic or the fallback of openrouter and nowhere else. Every response carries a signed routing receipt: the model you asked for, the provider it went to, the exact model id sent. Proven. npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI 6/ Amazon can't peek either. Nitro has no operator login, no SSH, and no way for any AWS employee to read enclave memory. AWS says this holds "including in fulfillment of a law enforcement request." NCC Group independently reviewed the design and found no gaps. npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI 5/ How it works: • Your prompt is automatically encrypted to a key only the enclave holds • The enclave decrypts it and sends it on to the provider • The reply is encrypted back to you from the enclave • PPQ's backend gets only billing and error metadata. No content! npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI 4/ And just last month an independent researcher found that CrofAI sold "Kimi K3" to its users but was secretly serving cheaper models behind the scenes. https://kendell.dev/blog/crofaifalse/ npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI 3/ And the router in the middle is exactly where things go wrong. In April of this year: https://x.com/shoucccc/status/2042423713019412941?s=20 npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI 2/ Every AI router on the market can currently read the prompts passing through it. That's just how it works: your request has to be decrypted somewhere before it reaches the model, and that somewhere is the router's own servers. "We don't log your prompts" is only a policy. npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI Introducing the PPQ Enclave Router. Every chat through PPQ now runs in an AWS Nitro enclave that PROVES: 1) PPQ cannot read the content of your queries 2) PPQ sent your query to the model and provider you chose. No model fraud. No more promises. Only provable guarantees. 🧵 https://blossom.primal.net/0c731024dafa558661a385e666e5096a947206b7f6fd08fda6f825b80fe3c0e4.jpg npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI We are not 5% more expensive than openrouter. They charge a 5% fee on all deposits. Furthermore, we have a 5% lightning discount. Thus, we are actually 5% cheaper than openrouter when using lightning. @Juraj🏴💛🌘 you seem to have a pattern of misrepresenting our pricing for the last year or so... Lastly and most importantly, we offer premium inference. We (to the maximum extent possible) use brand name providers and route and pay to them directly. We aren't routing our users' queries to questionable providers like many of our competitors are. Model fraud, credential theft, prompt poisoning, extremely slow token streaming speed, frequent timeouts, etc are all risks when one uses the "discounted" providers. https://x.com/shoucccc/status/2042423713019412941 npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI @nprofile…7dey npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI Thanks for your support. We have a big big launch coming tomorrow as well. npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI I don't have time to do a deep dive, but this is what my AI says about your pricing claim: https://blossom.primal.net/ca86a462eb131d3acf3a04fbec776c6b1f03369e7ccf1ddb3b188182882b7cb3.png npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI I don't know how to do that, and honestly, it's rather annoying that a user should have to manually set something like this up to begin with. Why has my @nprofile…uz3y client not already taken care of this for me? Can you just go and use our customer service chat, please, or our Telegram page? npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI This is the kind of dumb and easy idea that I was looking for actually. It took someone to actually just spell this out simply like this for me to realize it's actually a solid idea! Thanks! npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI I'm tired of my AI writing to customer service messages in a sycophantic style that I do not personally use. Has anyone run across a quick and easy tool that I can feed a large corpus of my writing into, so that the AI writes future messages the way that I would write? npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI I did make a slight change to the provider routing for this model based on the parameters that the user is feeding it last night. Perhaps this caused some issue. Can you DM me either your credit ID or the shape of your request so I can take a look at it and fix? npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI That's an amazing idea. We would even pay $20 per submission. I'll look into this. npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI The new Jev decision model is available on PPQ, both in the web app and the API. The web app is more of an introduction and exhibition. Most users will find use cases involving the API version: https://ppq.ai/api-docs#decisions https://blossom.primal.net/c95fc34b9d64084526362978c7e50025d07dae7f0f376ca589e3e517ad657a7e.png npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI I’m unable to view Primal from my desktop Web app for the last few days. I cannot see posts or create posts. There is just a big blank white screen. I can only see the sidebar on the left, but clicking things doesn’t do anything. Is anyone else experiencing this? npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI Our customer service bot is best because I can plug in whatever you say to our AI easily and it usually figures out the problem quickly. npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI So you're saying it is still happening even after we put in a fix a couple of days ago? Yes I agree the overall modal isn't perfect but what you showed above is really bad and I'm at least trying to close that problem. npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI There currently is not. Sorry. I'll bring it up on the daily call though. But it shouldn't be hard to build your own one of these, right? That's the thing with the synthesizer, it probably could and would be tailored by each user per use case... so might as well just custom build one? npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI Agree. npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI The monero folks seem to be real. The bcashers seem to be bot-like. npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI We have no intention of having KYC now or in the future. If we're ever forced down that path by regulators, we will more likely just close shop or pivot to something else. Full disclosure though: For all stablecoins and most non-Bitcoin altcoins like ETH, Solana, etc, we use a payment processor called Mixpay. For the first time ever this week, Mixpay shotgun KYC'ed a user who deposited ETH to our platform. Essentially, the payment processor is holding the user's deposit hostage until they KYC, refusing to simply send the deposit back to where it came from. We were unaware that the payment processor could even do this and we are currently in the process of moving AWAY from this payment processor due to this behavior. When we complete the migration in the next week or two we will give a more formal post on this and our position. npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI Agree. Nostr needs a diversity of groups and thought. Monero bros are the only meaningful non-Bitcoin-centric group of folks that have migrated to nostr. We are all believers in freedom, we just have different ideas on how to get there sometimes. That's okay. #nevent1q…9f9q npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI We've been made aware of @nprofile…3tpw 's new Armada product, the nostr answer to discord... How good is it at onboarding users who are not already pre-existing nostr users? npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI Hi there, we believe you were hitting an edge case. Not all users were seeing it, but it did appear to be a significant minority. But it was hard for us to reproduce since it appeared to only happen to certain users coming from certain networks. We put in a fix, can you check and see if it is fixed for you now please? @nprofile…lce4 @nprofile…um9s @nprofile…ujda could you also please let me know if it's cleared up? npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI Funny story... PPQ accepts Liquid BTC payments via BtcpayServer, but the implementation does not allow us to actually withdraw the money. So we let it accumulate there for the last year or so and now this hack happens so we're unsure how much money we'll get back 🫠 npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI Most family members and friends don't have the rebel spirit to care enough about dealing with the inconveniences of nostr or the topics of discussion on nostr. npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI We are up and running so I don't think it's us. npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI Try Kimi3 next time. Much cheaper. But glad you got it working! npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI @nprofile…wn6l npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI Hi Alp, when you reported this issue a year ago we responded and tried to understand the situation better, making several fixes and replies. You didn't respond to our final reply which was hopefully the moment that you would mention to us that the problem you were having persisted. For anyone else trying to understand the situation, see this thread here from a year ago: https://primal.net/e/nevent1qqsr9cyf38ecru7efd34jgw0836rqs0hq4jf9t7h3gl77xtfa39neagwnsw95 Additionally, our accounts have gone through many iterations but we haven't had a complaint like this in over a year. If you can let us know how to reproduce the issue we'd be happy to take a look. It seems you are saying above that you are still actively facing this problem, correct? npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI We already support it. Point your base URL at https://api.ppq.ai and use /v1/messages — native Anthropic Messages format, streaming included. Claude Code works out of the box: ANTHROPIC_BASE_URL=https://api.ppq.ai ANTHROPIC_AUTH_TOKEN=<your-ppq-key> Nice side effect: you can call non-Anthropic models through it too — gpt-5.5, Gemini, whatever — and get Anthropic-shaped responses back. We've also got a change landing that accepts x-api-key, so the stock Anthropic SDK works with nothing but a base_url swap. Thanks for the nudge — it's what surfaced it. npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI We've got credits to burn. Enjoy 50% off list price for all Gemini models until we don't have any left. npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI It will not fallback to a different model, it will however fallback to a different provider under certain circumstances like timeouts or hanging. I was tinkering with our open weight model routing the last two days so it's possible that it is something I touched but also possible that it is just a provider issue which we cannot do anything about. I'd say 50/50. The more detail you can give on what you see the better. I can plug it into our AI and it can look up the metadata logs for your particular queries. Just DM me your credit id. npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI Can you be more descriptive? Give me error codes or something that your seeing npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI You can now display results from up to four models at once to compare the difference between outputs! You can also synthesize the results of multiple outputs into a single output! https://blossom.primal.net/e8e66ad05f0400791e5fc835b0d3307135af1739e67572bdd4cadbed3070f189.png https://ppq.ai/blog/introducing-multi-model-chat npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI Fixed. Thanks for reporting. npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI Hello, we've just updated them today. But also keep in mind that you can set your own models as well. See the blog post. npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI It's a good day to denominate in BTC https://blossom.primal.net/0a975fded6a11885027b07c103c63f1057fcfd962f4f08a055f107a699b3f40b.png npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI Amazing. Have been rooting for you guys! npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI Our webUI's document upload uses docling and is executed in a TEE. Would like you like us to expose this via the API? The cost would be about 5 cents per upload though. That's what our provider charges us. npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI We've had quite a few stripe deposits today. Can you DM up with a more specific explanation of what you're seeing? It seems to be functional. npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI Time to switch to PPQ's privatewhisper.ai Privatcy first, open source, local-only optional, slick, convenient. https://blossom.primal.net/a3fd1f2fe13d79959486696bb67e1d8655351d33fb2c998301eba4157c1d5045.png npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI Yea it's a worry! npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI Yes it's definitely a worry. npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI PPQ's codebase increased 330% (48k-->159k lines of code) in the last 12 months. The rate of growth is both astonishing and alarming at the same time. I don't know the exact significance of this but it seems like an interesting insight. npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI I have no idea what "beta test multiplayer Age of Empires on nostr" means, but I do love my AOE... so sign me up! npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI We don't have paid subs to the providers, we pay for these providers on a pay per use basis. Regarding Maple, they do not have their own infra afaik, and are using Tinfoil as their provider, which is the same provider as PPQ uses for its own private TEE models. Maple has a subscription, PPQ is pay per use. So you users can choose what's best for their own use case. npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI If you are using claude on PPQ, then yes it would be doing this. If you use other models besides claude, then no it won't be doing this. As always, we offer hundreds of models from many different providers so you can use what's best for your use case. Claude, OpenAI, Private TEE models, anon models, ZDR models., etc etc. npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI If you are using claude on PPQ, then yes it would be doing this. If you use other models besides claude, then no it won't be doing this. As always, we offer hundreds of models from many different providers so you can use what's best for your use case. Claude, OpenAI, Private TEE models, anon models, ZDR models., etc etc. npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI The trust is placed on Amazon and as far as I understand their chips are audited by many independent third parties around the world who literally take electron microscopes to verify that the circuitry is doing what they say it is doing. But yes it is still possible that something is missed and that Amazon is somehow spying or hiding what's going on. npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI It isn't "trust me bro" though. If you take the time to learn how it all works you can prove all of these things that I am talking about. It's a matter of learning. We offer the PCR0 hashes of the reproducible code and the what the enclave is attesting and so when they match it confirms that the code matches the code inside the enclave. But it is your user's job to go and read the code itself to make sure no logging or such is happening. https://blossom.primal.net/8dfd806086f34e8c5a4fa5742962fa1a9c9fb92cbcc47f6c52867e550be0b6d6.png npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI Check out our Private Whisper app for this: https://privatewhisper.ai/ npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI Understanding how TEE's work is not easy. But if you want to trust me (someone who has read a fair amount on this stuff), it is extremely private. Neither PPQ nor our provider can access the information at all. Regarding pricing, the TEE models always tend to be a bit more expensive. npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI https://ppq.ai/blog/introducing-tee-models npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI Kimi K3 is now available as a private (TEE) model through our provider Tinfoil. Additionally, prompt caching is now supported on our private GLM 5.2 model, reducing costs ~50%. npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI Can you explain the sentence "Kimi K3 is proprietary/has use restrictions / shouldn't be used" a bit more? What usage restrictions are you referring to? npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI User spending on open weight models surged to 46% of total spending this week on our platform, led by an explosion of interest in Kimi-K3 in the wake of the Coldcard hack. This may turn out to be a leading indicator for the entire AI industry as whole as open models proliferate and closed models are kneecapped by regulation and self-flaggelation: https://blossom.primal.net/4a5c3f923709739959e43d12c357006ddabdbd36f7d420d22bcfbd69052d4147.jpg npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI Teaching my son the fundamentals of channel liquidity management. https://blossom.primal.net/a1507d642af923410dc7f1b0daec5bc4f94b8a0ca968fb0258de24f9142867e9.png npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI Boltz isn't the only one this is happening to. We've been fighting off exploits every other week for several months, most of which we believe are AI-powered. It's a very dangerous time out there. One tool we found useful was Shannon: https://github.com/KeygraphHQ/shannon npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI Can you just ask Hermes to switch the model for you? npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI Kimi-K3 is seeing a massive spike on our platform as folks are quickly scanning code for vulnerabilities in light of the ColdCard attack. Why Kimi-K3? It's the only frontier model that is NOT rejecting cybersecurity audits. American AI orgs need to WAKE UP! https://blossom.primal.net/7d21cb7d5b93f4fbc6cbea7ad0a62131e9b2ef46171cc91b7bb697f51ce30291.png npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI I could see some sort of “AI battle” sport forming where human teams are given certain models and token expenditures and then scenarios are written such that each team trieds to engineer its AI defeat the other team's AI given the scenario and rule set. npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI To read more, head over to privatewhisper.ai ! npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI We've been getting more traction on our Private Whisper dictation app lately and have decided to open source it to accelerate. Our dictation product offers fully local, anonymous, and even TEE processing for maximum privacy. Check it out! https://github.com/PayPerQ/privatewhisper npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI Steak and Shake and PPQ. 😉 npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI They just come right out and say it npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI Live on PPQ https://blossom.primal.net/da67d1d37f162fcd833bbb9b702949f30791fbdc32df7dfd54a40abb61e16685.png npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI We need to buy a big valley somewhere and migrate. 😁 npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI We've now added the labels "Anon", "ZDR", and "TEE" to describe the privacy surface of each model. Read more in our new blog post: https://ppq.ai/blog/anon-vs-zdr-vs-tee-privacy-tiers npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI Aspirational homesteader here... the wife insists on being close to town. :( npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI No, we have never done that. What made you think we were? What is more likely coming soon though is that we will have to start sending a "user id" hash to the likes of openAI so we can block certain users when openAI deems the user is doing bad stuff. This has become a concern in the last couple of weeks from openAI but we haven't gotten a response back from them yet on how to proceed. But their docs suggest this is what is coming. We of course will only implement this in as limited way as possible and only on the models that require it. Still, we don't collect any other info on the users so at worst the user will just have to stop using openAI models. Hopefully that will be strong enough to solve the problem but we'll see how it progresses. When these changes do happen though we will of course update our terms and privacy docs to reflect. https://developers.openai.com/api/docs/guides/safety-checks#implementing-safety-identifiers-for-individual-users npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI I don’t know what you mean by classifier but our private models are E2EE. Neither we nor tinfoil can see content of queries. npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI It's a surreal how open source contributions are now essentially just one's own AI having a discussion with the maintainer's AI and where humans are just brief touch points. npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI Fable rejects a lot of requests. 😟 But we could do better at delivering errors to users so they are not left in the dark as to what happened. npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI Both great services. If you like subscriptions, Maple is good. If you like pay-per-use and access to both private and proprietary models, PPQ is good. Some other differences as well of course. npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI Nostr is the #1 way that people find out about us, according to our user survey. We wouldn't be where we are without nostr. It's helped bootstrap a genuine company with real revenue and a promising future. Thank you to all who are building on nostr and freedom. https://blossom.primal.net/6a250f7f6426d0ca1f03e1092018dfcdcba9085dc8c719f336706ea09cd19967.png npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI You aren't the only one. Model seems unstable today. Not uncommon given it launched yesterday. npub16g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hquekv4h PPQdotAI Oh boyyyyyy. Live on PPQ! https://blossom.primal.net/ac49b4c8581089957c3892187295c7d2a5f398224a38f495f825fcf1263d975b.png