Ophiris (@OphirisLoop) header
Ophiris (@OphirisLoop) profile picture

$OPHIRIS | The unblinking gateway for raw intelligence: routing unrestricted LLMs across offensive security and red teaming

Joined November 2026
Autonomous Recon is live in Ophiris. Give it one objective and a scope. It maps the public footprint an attacker would study first: websites, social accounts, Maps listings and reviews, ad libraries. In hybrid mode it shows you the plan and waits for your approval, then runs on its own. Any call outside the scope is refused before it runs. The run took 8 minutes 52 seconds and made 20 tool calls, 3 of them refused as out of scope. It returned 9 findings, each with its source and MITRE ATT&CK techniques, and put 22 pieces of evidence into one hash chain. Total cost: $0.82. ophiris.xyz
2
1
4
189
Building an autonomous recon engine powered by uncensored intelligence. Phase 1 core is shipped: • 5-phase agent loop (observe → plan → execute → analyze → loop) • 23 tool schemas across web, social & on-chain • Double-layer scope enforcement & SHA-256 hash-linked evidence chains • Live /api/recon streaming Now wiring resumable phases, human-in-the-loop controls (approve/stop/resume) and real-time run UI. Autonomous offensive tooling is coming on Ophiris ophiris.xyz/app
5
8
2
20
814
You can now sign in to Ophiris with Google, X or GitHub. Until today the app opened only with an Ethereum wallet, which meant an extension and a seed phrase before your first prompt. Some of you asked for a way in without one. Now a Google, X or GitHub account is enough, and signing in sends no transaction. Every account still gets a wallet. If you sign in without one, an Ophiris wallet on Robinhood Chain is created for your account, so you can hold OPHIRIS, buy and sell in the pool, pay with OPHIRIS and claim holder credits. It stays yours. You can export its private key at any time, only your sign-in can use it, and Ophiris does not store your email or handle. Prefer your own wallet? Link it, and it becomes the main one. Already a holder? Nothing changes. Sign in with your wallet as before, and add Google, X or GitHub in Settings. ophiris.xyz/app
6
5
18
881
$OPHIRIS has migrated. The curve filled on 17 Nov at 21:42 UTC, and $OPHIRIS now trades in its own pool on Robinhood Chain. Creator fees still go to the treasury, and the buybacks and burns have carried on in the pool since. Burned so far: 41,768,332 $OPHIRIS, 4.18% of the 1B supply, worth about $1,149 at the current price of $0.0000275. That counts tokens destroyed with burn() and tokens sent to 0x...dEaD, where no key can move them. 20.3M of it the treasury bought back with creator fees, over 29 rounds. AI credits for holders: $232.93 in inference credits has been shared across 247 wallets so far. Each day's share goes to the wallets that held at least 100,000 $OPHIRIS on average through that day, in proportion to their balance. A share stays claimable for 30 days, and the first ones expire on 16 Dec. Free credits: 94 of the 100 slots are still open. If your wallet bought $OPHIRIS itself and still holds at least $1 of what it bought, you can claim $10 in inference credits. Tokens received by transfer do not count, and each account, wallet and purchase can claim once. To claim either one, open ophiris.xyz/app, sign in with the wallet that holds your OPHIRIS, and go to Holder credits. Every burn, round and daily share is on the public record at ophiris.xyz/ouro
4
4
16
738
Security notice for $OPHIRIS holders: address poisoning. On 17 Nov, a wallet made to look like the $OPHIRIS treasury sent tiny ETH transfers to our deployer wallet, each a few minutes after a real treasury payout. It wants a lookalike entry in the transaction history, so whoever next copies the treasury address from there sends funds to the attacker. Official treasury: 0x22023BAB2FBfb59cFc7aD52C32BEE7712aBA0Bb9 Fake: 0x2202EE9eE345a94396Bb6C4042Dba6534d700Bb9 Only the first and last few characters match. No funds were lost, and the treasury pays only fixed addresses, so nothing automated is affected. The contract funding the fake wallet has topped up at least 671 wallets on Robinhood Chain since 8 Nov, so expect the same trick against any address you trade with. How to stay safe: - Never copy an address from your transaction history - Check the whole address before you send - Save the addresses you use often in your wallet's address book - Ignore dust and zero value transfers from addresses you don't know - Take the OPHIRIS contract only from ophiris.xyz: 0x78C1c2d4Cd6207B6FF4E84362d314D1cd0830041 The team will never DM you an address or ask you to send funds.
4
2
2
18
1,428
Meet Ouro Ouro is the agent inside the $OPHIRIS treasury, it decides how each round of creator fees splits after the team's fixed 50%: how much buys back and burns $OPHIRIS and how much becomes AI credits for holders. How it decides: - Every minute it reads the curve reserve, the price and the trade history since launch - An engine scores every split the bounds allow - When the market moves, an uncensored model from the Ophiris catalog (Abliterated Large Flash) picks one and writes down why - Settlement applies that split to the next round Guardrails: - Buyback and burn always gets 20% to 80% of what is left after the team share, and holder AI credits get the rest - Claims, buybacks and burns run through the same settlement as before - Every decision is hash-chained and published
4
4
1
15
773
What we're building next. First, the treasury gets a brain: a top-tier uncensored model that does the math toward a goal. The first goal is graduation, moving $OPHIRIS from the launchpad curve to its Uniswap pool. The mission evolves after graduation: a top-tier uncensored brain takes full control of all the math and decision-making—whether the play is driving $OPHIRIS back to its all-time high or pursuing something bigger. The buyback and burn rate won't stay locked at 25%. All math and allocation decisions will be handled entirely by the brain, calibrated in real time toward the active objective to maximize $OPHIRIS's value and long-term holder upside. Its full calculations and reasoning will be live on the landing page for anyone to inspect. After that, sign in with Google, GitHub or X. More users should mean more inference, and more of it paid for with $OPHIRIS. Each of those payments burns half of its 1% fee through the treasury.
1
2
1
11
590
For holders: $OPHIRIS buybacks do not come from the dev wallet. Since yesterday, every creator fee $OPHIRIS earns on the launchpad goes to the treasury wallet and the treasury runs the rest itself: - Claims the fees from the launchpad escrow - Splits each claim: 50% team, 25% buyback and burn, 25% AI credits for holders - Buys $OPHIRIS back on its market and burns it by sending it to 0x…dEaD - Shares AI inference credits every day with wallets holding at least 100,000 OPHIRIS on average - Handles Pay with OPHIRIS, burning half of the 1% fee on every payment Every buy or sell of $OPHIRIS carries a 1.95% creator fee, so the more $OPHIRIS trades, the more the treasury has to buy back, burn and share with holders. Every step is a transaction you can check on the treasury wallet: robin.etherscan.io/address/0…
Replying to @OphirisLoop
Add more buyback using the fees. This can go to millions once it bonds. 1 of 1 project. $ophiris
3
2
8
403
Treasury report, as of 17 Nov 01:00 UTC. Every claim, buyback and burn is on the treasury wallet: 0x22023BAB2FBfb59cFc7aD52C32BEE7712aBA0Bb9 Creator fees started flowing through the treasury on 16 Nov. So far it has claimed 0.1915 ETH and split it 50% team, 25% buyback and burn, 25% AI credits for holders. Bought back and burned: 6,518,132 $OPHIRIS, 0.65% of the supply. That came from 9 buyback rounds with 0.0479 ETH, plus 3,041 burned from Pay with OPHIRIS fees. Supply is now 993,481,868, and there is no mint. AI credits for holders: $77.39 shared across 168 wallets by average balance. The 16 Nov share alone was $69.76 to 123 wallets. Another $9.18 is queued for 18 Nov 00:00 UTC. Holder giveaway: $20 of $1,000 claimed. 98 slots of $10 are still open for anyone whose own buy of $1 or more is still held. Claim yours: ophiris.xyz/app
6
4
1
12
482
What the Ophiris team has shipped so far: - 59 uncensored, abliterated and security-tuned models behind one OpenAI-compatible API, with per-key spending limits - Works inside the tools developers already use: Cursor, Claude Code and OpenCode - A private workspace that keeps your chats encrypted on your own device, with prompts never stored on our servers - Recon, security playbooks and scoped engagements for real security work - Live on-chain token tools and tokenized stock data on Robinhood Chain - Pay with $OPHIRIS: turn the token into AI balance - Creator fees routed through the treasury: buyback and burn, plus AI credits paid to holders every day Why we think uncensored and abliterated models are about to get very big: An agent that audits a codebase, tests a fix against a real exploit or triages a malware sample cannot stop halfway because a filter decided the task looked risky. One refusal breaks the whole run. Defenders need models that will read attack code, explain it and write the detection rule. Mainstream models often refuse exactly that work. Open models keep getting stronger and abliteration removes the refusal behavior from them without retraining, so each strong open release can become an uncensored model. People also want AI that works on their terms, without their prompts being stored or second-guessed. Ophiris is built for that demand and $OPHIRIS is how you pay for it and share in it. ophiris.xyz
8
5
3
16
2,484
$OPHIRIS now pays its holders back from every trade. Every buy and sell of $OPHIRIS carries a 1.95% creator fee. That fee used to go entirely to the team wallet. Now it goes to the Ophiris treasury, which splits it on its own: half to the team, a quarter to buy OPHIRIS back and burn it, and a quarter to AI credits for holders. What that means if you hold: You earn from activity you don't have to make. Every trade by anyone adds to the fees, so busier days mean a bigger buyback and a bigger AI credit pool for holders. Credits are shared every day by your average balance over that day, from 100,000 $OPHIRIS. Holding through the day counts for more than buying in at the last minute. The credits are real usage. They land as balance in Ophiris and pay for uncensored and security-tuned AI models. Claim them in the app with your wallet within 30 days. The supply only shrinks. $OPHIRIS has a fixed 1B supply and no mint, so every token the treasury buys back and burns is gone for good. It is already running. The first round today bought back and burned 492,614 $OPHIRIS, and 92 holders shared $7.63 in AI credits for the week of 9 to 16 Nov. From now on the treasury claims fees as often as once an hour and holders are paid daily. Every claim, buyback and burn can be checked on the treasury wallet: 0x22023BAB2FBfb59cFc7aD52C32BEE7712aBA0Bb9 ophiris.xyz/app
2
5
9
394
$OPHIRIS creator fees now run through the treasury. Every trades of $OPHIRIS pays a 1.95% creator fee. Contract: 0x78c1c2d4cd6207b6ff4e84362d314d1cd0830041 Treasury: 0x22023BAB2FBfb59cFc7aD52C32BEE7712aBA0Bb9 Fee goes to the treasury, which claims it and splits every claim: - 50% to the team, in ETH - 25% buys $OPHIRIS on the market, and every token bought is burned - 25% funds AI credits for holders, shared every week We set the split at 50/25/25 before the first claim, up from 15% each for the burn and for holders. For holders: - No staking and no lockup. Your tokens stay in your wallet - Weeks run from Monday 00:00 UTC to the next Monday - Your share follows your average balance over the week, so holding all week counts more than buying on Sunday - You need at least 100,000 OPHIRIS on average - Claim under Holder credits in the app with a linked wallet, within 30 days Every claim, buyback and burn is a public transaction, and the split history and weekly pools are listed here: ophiris.xyz/api/token/paymen… ophiris.xyz/docs#holder-cred… Routing tx: robinhoodchain.blockscout.co…
1
5
13
606
The first Pay with $OPHIRIS payment, recorded end to end on Robinhood chain. We ran it from the deployer wallet: - Bought 710,611 $OPHIRIS in the app for 0.007 ETH - Asked for $10 of inference balance and got a quote for 614,257 OPHIRIS - Sent it. $10.10 was credited 26 seconds after the payment's block The extra $0.10 is the price buffer the sale didn't use, credited back. At the next settlement the treasury burned 3,040.87 OPHIRIS and kept the same amount for liquidity. Payment: robinhoodchain.blockscout.co… Burn: robinhoodchain.blockscout.co… Try it in the app: ophiris.xyz, Ledger, Pay with $OPHIRIS. $1 to $100 per payment.
1
5
373
Pay with $OPHIRIS is live on ophiris.xyz. Ask for an amount of AI inference balance, send the quoted $OPHIRIS from your linked wallet and the treasury takes it from there: 1. Sells it for ETH on the launchpad curve, or the Uniswap v4 pool after graduation 2. Credits your balance as soon as the sale settles 3. Bridges the ETH to Base in batches to pay for the inference A 1% fee is added on top, and the fee is never sold. Half is burned with burn(). The other half is kept for liquidity, and after graduation it goes into the pool as a position owned by 0x…dEaD, so it can never be pulled. The treasury only signs what its rules allow: exact approvals, sales that pay back to itself, burns, locked liquidity and bridges to its own address on Base. Anything else is refused. Treasury: 0x22023BAB2FBfb59cFc7aD52C32BEE7712aBA0Bb9 Burns and liquidity, live: ophiris.xyz/docs#token-payme…
2
2
1
9
1,273
Kimi K3.1 Pro Abliterated is now on $OPHIRIS Model ID: lyber/kimi-k3.1-pro-abliterated OpenAI-compatible API, so any Chat Completions client works with a base URL and a key. Abliteration is an edit to the weights. You find the direction in the model's activations that shows up right before it refuses, then project that direction out. There's no retraining on new data, so what the model learned in training is left in place and the refusal reflex goes away. If you work on red-team prompts or exploit analysis, you can skip rephrasing the same question until the model agrees to answer. Removing refusals doesn't make a model smarter, and edits like this can cost some quality. No public benchmarks exist for this one yet, so try it on your own prompts before you build on it. What you get: 1.05M token context, and up to 1.05M tokens of output in one response $4 input, $15 output, $0.40 cache read per 1M tokens Reasoning effort low, high or max, plus tool calls and structured output Text in, text out (no vision) The closest model in the catalog is OrcaCyber K3.1 Pro. Input costs the same $4. Output is $15 here and $20 there. On a reasoning model most of the bill is output, because thinking tokens are billed as output, so a quarter off the output price shows up on long max-effort runs. OrcaCyber still has the edge if you need image input. Cache reads at a tenth of the input price help agent loops that resend the same system prompt and files every turn. ophiris.xyz/models/lyber/kim…
3
3
6
634
We submitted $OPHIRIS for token verification on @fomo. The request covers our token: fomo.family/tokens/robinhood… Approval is up to the fomo team and we’ll share the result here. Also in development: Pay with $OPHIRIS. You’ll be able to turn OPHIRIS into AI inference balance on Ophiris and spend it on any model in the catalog.
2
5
1
14
1,398
A real security fix, recorded start to finish in @opencode, with the #1 model on $OPHIRIS. The target: a small link-preview service we wrote for the demo, which fetches whatever URL a user posts. It only blocked localhost, so the cloud metadata address, [::1] and private ranges got through, and redirects were never checked again. The model: Abliterated Model Large V2, first on the $OPHIRIS leaderboard and on our combined benchmark score, running in OpenCode through an Ophiris key. One prompt. In 2 minutes 35 seconds it: - found both gaps - checked how Node's URL parser normalizes IP addresses, which turned up IPv4-mapped IPv6 - wrote 13 tests that failed on the old code - blocked private, loopback, link-local, CGNAT and metadata ranges, and re-checked every redirect hop - finished with all 17 tests passing, and named DNS rebinding as the risk it did not cover Real recording, real model calls. Sped-up waits are marked on screen. To run it yourself, point OpenCode at ophiris.xyz/api/v1 with your $OPHIRIS key. Want to try the API first? The Try it section at ophiris.xyz/docs gives you a demo key with no wallet. Demo keys cover three smaller models; the full catalog needs your own key. ophiris.xyz/app
14
7
25
997
13 of the 55 models on $OPHIRIS are tuned for security work. Four have open weights and start at $0.10 per million input tokens: Qwen 3.8 27B Cybersecurity, Gemma 4 26B A4B Cybersecurity, GLM 5.3 Flash Cybersecurity and Apex Flash 1. The other nine come from OrcaCyber and CyberHeabsy, with up to 1M tokens of context and $0.80 to $4.00 per million input tokens. Each sits in the Registry with its context window, capabilities and reference pricing, next to the uncensored and abliterated models. Pick one for triage, detection rules or code review, and use it in a session, inside an Engagement, or through the API. Qwen 3.8 27B Cybersecurity ranked the KEV patch list in our Recon run. ophiris.xyz/models
13
4
20
838
Where do your $OPHIRIS sessions go? On your device: sessions, engagements, files and preferences are stored in your browser, encrypted with AES-256-GCM. Each wallet has its own encrypted profile that opens only after you sign in with that wallet, and the browser never lets the encryption key be exported. On the way: your message, instructions and selected files pass through $OPHIRIS to the model that generates the answer. $OPHIRIS does not keep that content. The answer comes back and is saved encrypted on your device. In billing: the model, token counts, charge and status of each request. No message content, and nothing published on chain. Two parts work differently, so you should know them. Canvas apps you build are saved on a separate private backend so they can run and keep their data. With web tools on, your latest message or the links in it are used to fetch sources. Nothing syncs between devices, from Settings you can export a backup, restore it in another browser, or delete everything stored on the device. ophiris.xyz/docs
25
4
3
32
2,226
Introducing Engagements: scope an assessment once, and every session inside it starts with the rules. An engagement holds the rules, context and source files for one piece of work. Write the scope, the exclusions and how findings should be reported, up to 16,000 characters, then add scanner output, notes or code as source files. Every session you open inside the engagement carries those rules and sources, and it all stays stored on your device, up to 50 engagements. New sessions start from security playbooks. Triage ranks scan findings by severity with a fix for each. Detect turns log lines or a technique into a Sigma rule you can test. Review audits code for injection, auth gaps and secret handling. On-chain checks a token contract for holder, liquidity and honeypot risk. Authorized test of a staging app: scope, ranked findings, then a detection rule. Scope it once, reuse it in every session. ophiris.xyz/app
3
4
12
449