I kept coming back to one line from a builder note that landed early on October 3, 2026. Building, the writer argued, had finally gotten easier. The hard part was no longer the chain. It was choosing what to ship, and then getting a stranger to care. That sounds almost too plain for crypto, which usually prefers a louder slogan. Still, if you have watched teams grind for months on a protocol that nobody outside the Discord ever touches, the sentence lands. Illia Polosukhin, co-founder of NEAR, used that framing to drop a loose stack of product sketches: AI pets that carry their own compute budget, private trades that do not require you to already know the other side, and payments that hop chains without a tutorial. Thirteen application ideas, plus a handful of distribution hints. Not a roadmap. More like a dare.
Why Product Choice Matters More Than Another Protocol Demo
Polosukhin’s note was aimed at people who had started, or restarted, work on NEAR over the previous couple of weeks. The claim was not that the chain had suddenly become magic. It was narrower. Tooling had thinned out the usual excuses. Once the compile-deploy-pray loop stops eating a weekend, you are left with taste. What should exist that does not already exist in a half-finished form on three other networks?
I have found that lists like this age badly when they read like a pitch deck. This one is more useful if you treat it as a map of friction. Every idea points at a place where crypto still asks too much of a normal person. Pay a merchant without learning a new wallet. Show up to an event without trusting a spreadsheet. Let an AI character burn compute without handing a company your card. Trade size without broadcasting the whole book. Those are old complaints. The interesting part is how he tried to pin each one to infrastructure that already exists, rather than to a chain that still needs to be invented.
Building itself got easier too, so it’s really about what to build and how stand out to get distribution.
Illia Polosukhin, October 3, 2026
Distribution sits in that sentence on purpose. A clever contract with no audience is a hobby. He named communities, launch boards, and cross-promotion as the unglamorous half of the job. I will come back to that, because the product ideas are the part people will screenshot, and the audience problem is the part that usually kills them.
A Checkout Button That Does Not Care Which Chain You Live On
The payment sketch is the most merchant-shaped of the bunch. A universal checkout widget, offered to shops, that lets a customer pay through NEAR Intents using the wallet and the chain they already hold assets on. The developer’s job, in this telling, is not to convince the buyer to bridge. It is to hand the merchant a button, then push that button toward larger storefront software. Shopify was named as a distribution target, not as a signed partnership. That distinction matters. A target is a wish with a logo on it.
Still, the wish is coherent. Most checkout experiments in crypto die in the same ditch. The buyer has funds. The merchant wants dollars, or at least something they can reconcile on Monday. Between those two facts sits a bridge, a gas token, a new app, and a support ticket. Intents, in the version Polosukhin is selling, are meant to collapse that sequence into one signature. You state the outcome. Solvers and liquidity do the ugly middle.
Would I put this on a real shop tomorrow? Not without settlement guarantees a finance team can read without a translator. Perhaps the most interesting aspect is not the widget itself. It is the sales motion he implies. Do not wait for the merchant to discover the chain. Bring the button to them, then beg the platform that already has their theme files. That is a distribution strategy dressed up as a product idea, which is probably the right order.
- The buyer keeps the wallet they already trust.
- The merchant sees a completed payment, not a chain tutorial.
- The developer’s real work is integration and support, not a new token.
- Storefront platforms are the bottleneck, not the smart contract.
Event Deposits That Only Pay the People Who Walk In
Conference tickets are a quiet graveyard of good intentions. Someone registers, someone no-shows, the catering order was already placed, and the group chat argues about refunds for a week. Polosukhin’s version borrows from an older project called Kickback. Attendees lock a deposit in escrow. The host checks them in. People who actually appear then split the pooled funds. Miss the door, and your deposit feeds the people who did not.
The extension he wants is cross-chain. Same ritual, assets from more than one network, so a Solana-native guest and a NEAR-native guest are not forced through the same onboarding sermon. I like this more than the pet idea, if I am honest, because the incentive is rude and clear. Skin in the game beats a reminder email. It also creates a nasty edge case. What if the host never checks you in? What if the venue wifi dies and the signature fails at the door? Escrow without a dispute path is just a nicer way to lose money.
A workable version needs a timeout, a fallback checker, and a rule for partial attendance. Crypto events love optimism. Accountants do not. If someone ships this, the unsexy screen, the one that says “host did not confirm, funds return on Friday,” will matter more than the animated ticket.
Wallets That Visit Another Chain Without a Bridge Lecture
Distribution advice in the same note pointed at Aurora’s Intent Connect. The pitch is simple enough to repeat at a meetup. Let people use an application that lives on another chain without arranging a bridge or learning how NEAR feels under the hood. Context already exists for that claim. On September 17, Aurora Labs described one-signature execution into Sui applications, with supported assets entering those apps through Intents Connect. Users, the company said, could finish supported actions without switching wallets, bridging, or buying the destination gas token.
NEAR Intents, in that setup, supply liquidity, settlement, and the links between chains. The application keeps its own product surface, whether that is lending, trading, staking, or yield. I would not treat a press line as proof that every action is smooth. One-signature flows have a habit of hiding a solver fee inside the quote. Fine, if the quote is honest. Ugly, if the user only notices after the third try.
Even so, the design choice is the right argument against the old multi-chain tax. You should not have to become a part-time bridge operator to lend ten dollars. If Intent Connect holds up under real traffic, it is less a feature and more a permission slip for apps that refuse to pick a single home chain.
AI Pets With a Wallet, a Wardrobe, and a Compute Bill
Then there is the consumer bait. An “AI Tamagotchi,” characters represented as non-fungible tokens, each carrying a NEAR balance that pays for inference. NEAR AI would run verifiable inference from a prompt encoded onchain. Feed the creature. Dress it. Later, maybe arenas and competitions. It is easy to smirk. Digital pets have been a punchline since the late nineties, and crypto has already minted a museum of abandoned companions.
The mechanic underneath the joke is sharper than the costume. The pet is not subscribed to a company plan. It holds a balance. When the balance runs dry, the character goes quiet unless someone tops it up. That is a different relationship than a monthly SaaS charge you forget to cancel. Ownership of the token and ownership of the compute budget sit in the same object. Whether anyone wants that object is a separate question. I suspect a narrow group will, the same group that already names in-game items and argues about skins.
Related plumbing was already public before this note. In late July, NEAR described staking-based AI payments that turn locked tokens into monthly computing credits. At launch the system covered 43 models, including ones from OpenAI, Anthropic, and Google. Users kept ownership of the staked tokens and could recover them after unstaking. The locked balance set the credit line. A pet that “eats” from that credit line is mostly a skin on a billing system. Skins are how consumer crypto usually sneaks a serious primitive past people who would never read a staking doc.
Pet loop, stripped of the mascot: NFT identity + onchain prompt + NEAR balance for inference + verifiable run = a character that can go broke
Battle arenas are where this either becomes a game or becomes a farm. If winning prints more compute credit, you have built an incentive to script the pet. If winning is cosmetic, you have built a social toy. Both can work. Mixing them without saying which one you chose is how these projects get ugly.
An Encyclopedia That Refuses to Trust the Edit Button
A second information product in the note flips the usual wiki. Facts become ongoing prediction markets, in the spirit of Augur. Pages are not edited by whoever shows up with a keyboard. An AI drafts them from the market’s current read on each claim. I have mixed feelings here, and I think that mix is healthy.
Markets are good at pricing things people will bet on. They are shaky at things nobody is paid to care about, and worse at things a motivated group can bully. An encyclopedia built only from market odds will be loud on elections and sports, thin on everything else, and occasionally confident in the wrong direction because the liquidity was bored. Using AI as the writer does not fix that. It launders the odds into prose that sounds finished.
There is still a real product hiding in the sketch. A page that shows the claim, the market, the dissent, and the model’s summary, with the summary clearly labeled as a reflection of bets rather than a verdict. That is closer to a research terminal than a reference book. Call it an encyclopedia and readers will expect neutrality the mechanism cannot promise.
Testing Models Against Data You Refuse to Hand Over
The benchmarking idea is quieter and, to my eye, more serious. Test models against private datasets without exposing the underlying material. Certified results tied to a specific model and to the identifying hash of the test set. Developers submit a testing framework, or a privately hosted model. The point is a receipt. This model, this hidden set, this score, this proof, without the set walking out the door.
Companies already pay for private evals. They also already leak them, by accident or by a contractor’s laptop. A chain receipt does not stop a bad employee. It does give a buyer something sturdier than a PDF with a logo. If the hash matches and the execution proof holds, the argument shifts from “trust our benchmark team” to “check the run.” That is a better argument. It is also slower to explain in a sales call, which is why most eval shops will keep the PDF anyway.
| Idea | User pain | What has to be true |
| Checkout widget | Paying across chains | Quotes stay honest, merchants reconcile |
| Event escrow | No-shows and refund fights | Check-in cannot be gamed or skipped |
| AI pet | Compute bills and ownership | Balance, prompt, and proof stay linked |
| Private evals | Hidden test data | Hash plus certified run, no leakage |
| Deal marketplace | Finding a counterparty | Intents fill without doxxing size |
Private Deals That Do Not Start With a Phone Call
On trading, Polosukhin pointed at a known snag in NEAR Intents’ existing private deals. You currently need to know who you are trading with. That is fine for a desk that already has relationships. It is a dead end for everyone else. His fix is a marketplace where people post available deals and others fill them through Intents. Finding the counterparty becomes the product. Settlement stays in the intent layer.
Over-the-counter flow has always been a social problem pretending to be a technical one. Size wants privacy. Price wants competition. Those two wants fight. A board of anonymous intents can help, right up until the board itself leaks who is desperate. I would watch the metadata harder than the matching engine. Timestamps, size buckets, and repeat posters tell a story even when names are missing.
Social trading gets a related treatment. Sell viewing keys so selected customers can see otherwise private transactions. The trader can show profit and loss in public, while the activity a follower would copy stays behind the key. That is a cleaner split than the usual copy-trade feed, which shows everything or nothing. It also creates a new thing to steal. A viewing key is a password with a P and L attached. Lose it, and your book is a spectator sport.
Never Get Liquidated Is a Slogan, Not a Promise
One label in the note should be read with the safety on. “Never get liquidated.” The proposal is to use NEAR Intents to manage lending and perpetual futures positions, seeking interest while trying to keep liquidation at bay. He framed it as an application idea, not as a demonstrated guarantee. Good. Slogans that sound like insurance are how retail gets hurt.
A system can reduce liquidation risk. It cannot abolish gap risk, oracle lag, or a solver that fails at the worst minute. If a team ships this, the honest homepage says “we try to roll, hedge, and top up before the line is crossed,” not “never.” The difference is not tone. It is whether the user understands they can still wake up flat.
Beside that sits “delta neutral anything.” Spot holdings paired with perpetual futures, under a NEAR contract that routes transactions through Intents. The appeal is obvious. Hold the asset you want exposure to, short the perp, collect a funding or basis spread, and let the contract babysit the legs. Delta-neutral books are a profession, not a toggle. Funding flips. The hedge drifts. The intent fails on one leg and you are suddenly directional while you are asleep. A contract that manages both sides is useful. A contract that implies the spread is free money is a marketing problem waiting for a bad week.
- Post the intent, do not assume the fill.
- Keep both legs visible, including failed hedges.
- Cap size relative to solver liquidity, not to hope.
- Show liquidation distance in plain language.
- Treat “never” as a bug in the copy.
A Listed Fund, and What It Does Not Prove
U.S. investors already have a listed route to the token itself. On September 29, Bitwise announced a U.S. NEAR exchange-traded fund on NYSE Arca under the ticker NRR, with a 0.75 percent management fee. The firm said the fund holds NEAR directly and plans to stake its holdings, with rewards accruing through net asset value. Chief investment officer Matt Hougan, explaining the investment case, pointed at NEAR Intents as settlement infrastructure for AI agents handling payments and swaps.
An ETF is not a product-market fit trophy for pets or private desks. It is a wrapper. It tells you that a regulated vehicle can hold and stake the asset, and that at least one allocator thinks agent settlement is a story worth telling institutions. Builders sometimes confuse that with demand for their app. They are cousins, not twins. Capital can arrive for the token while the checkout button still has twelve users. I would rather see the widget in a real merchant admin panel than another ticker announcement. Both can be true in the same month. Only one means someone paid for something other than exposure.
Private Code Review, Private Charts, Private Transcripts
The privacy cluster in the note is where the AI pitch either earns its keep or becomes a brochure. For software teams, an agent that privately reviews private pull requests through NEAR AI, with verifiable execution. Code is the asset. Sending it to a generic chatbot already makes security teams flinch. A review path that keeps the diff sealed, then proves the run, is a feature a staff engineer can defend in a meeting. It still has to be faster than the senior who already knows the codebase. Speed loses and the agent becomes a novelty tab.
Healthcare gets a stricter version. A frontend encrypts medical data end to end before it reaches NEAR AI for a second opinion. Crypto and fiat payments sit beside custom prompts stored in NFTs, which a user could pick for different kinds of expertise. I am wary of anything that puts clinical language next to a token standard. A second opinion is not a collectible. If the NFT is only a container for a prompt and a payment rail, say that. If it is a speculative object attached to someone’s scan, walk away.
The design constraint is still right. Encrypt first. Send the ciphertext, or a form the model can use without the clinic losing custody of the raw file. Pair that with a payment that does not require the patient to learn a seed phrase in the waiting room. Fiat on one side, crypto on the other, is an admission that most people will not stake a token to ask about a lab result. Good admission.
Desktop transcription follows a similar instinct, modeled on tools in the Granola mold. NEAR AI processes the transcription. Records stay local. Backups are encrypted with the user’s own keys. As an alternative to a subscription, fund the service through NEAR staking. Local-first notes are having a moment because people are tired of meetings living on someone else’s server. Staking-as-billing is the crypto-native twist. It will feel elegant to holders and invisible, or hostile, to everyone else. Ship the local archive first. The funding method can be a setting.
Privacy features fail in the settings screen long before they fail in the proof system. If the default is “share,” the cryptography is a costume.
Cap Tables, Fundraising, and the Boring Tools Companies Actually Open
The note also includes a capitalization-table tool with related fundraising instruments, crypto payment acceptance, and banking support. This is the least viral item on the list and possibly the one a company would pay for. Cap tables are where startups already live in spreadsheets they are afraid to email. Adding crypto payment acceptance does not fix messy equity. It does give a treasury a way to take a payment without a three-day wire, if banking support is real and not a logo on a landing page.
I would split this product in two. One half is records: who owns what, what instrument converted, what the option pool did after the last hire. The other half is money movement. Teams that fuse them too early end up with a wallet attached to a legal document nobody’s counsel has read. Start with exportable records and a clear payment leg. Glamour can wait.
Five Ways He Thinks Builders Actually Get Seen
Product ideas without a path to users were the thing he said he wanted to fix. The distribution suggestions are plainer than the pets. Engage NEARLegion. Reach other communities through cross-promotion. Put the product on Product Hunt and similar discovery boards. Talk about it on X. Use Aurora’s connection layer so people from other chains can try the app without a bridge homework assignment.
None of that is secret. All of it is skipped by teams who would rather write another contract. Cross-promotion only works if you have something the other community wants, not a banner asking them to bridge. Launch boards reward a clear first session, not a white paper. Social posts reward a demo someone can quote. Legion-style groups reward people who show up before they ask for a retweet. I have watched the opposite pattern too many times. Ship on Thursday, post a thread on Friday, go quiet when the likes stall, blame the chain.
If the tooling really is easier, the excuse shrinks. A weekend prototype of the event escrow, tried at one meetup, will teach more than a month of architecture diagrams. The pet can be ugly. The checkout button can support two assets. The private review bot can handle one repository. Distribution starts when a stranger completes the action without you in the chat.
What a Builder Should Steal, and What to Leave on the Thread
Taken together, the sketches cluster. Payments want a button and an escrow rule. Trading wants a counterparty and a key. AI wants a balance, a proof, and a reason not to send raw data to a vendor. Distribution wants communities that already gather, plus a way for outsiders to arrive without a new gas token. That is a coherent stack even if several of the costumes are optional.
Steal the constraint, not the mascot. A character that pays for its own inference is a billing idea. A market that writes the encyclopedia is a labeling idea with risk attached. A perp manager that says “never” is a copy problem. A viewing key is a permission problem. Intents are the shared pipe, which means the pipe’s failure modes are now your failure modes. Solver downtime, quote slippage, and chain-specific quirks do not vanish because the button looks universal.
There is also a cultural tell in the list. It assumes developers are no longer blocked on deployment, so the status game moves to users. That assumption will be true for some teams and false for others. If your week is still spent on indexing and wallet edge cases, a Tamagotchi will not save you. Finish the path a single user can walk. Then make it legible to a second community. The note’s best line is the unpolished one about standing out. Everything else is a menu.
Would I build the pet? Only if the compute balance is the product and the face is a skin I can delete. Would I build the event deposit? Yes, for a single recurring meetup, with a timeout and a human override, before anyone says cross-chain. Would I trust a liquidation slogan? No. The useful work is in the boring promises: funds return, quotes match, records stay local, proofs check out. Crypto has enough creatures. It is short on checkout flows a merchant will leave enabled after the pilot ends.
Polosukhin did not claim these thirteen sketches are live products. He claimed the building season is open and the scarce skill is choice plus attention. That is a narrower boast than the industry usually makes, and it is easier to test. Pick one friction. Ship the smallest thing that removes it. Show the result somewhere people already look. If the AI pet never leaves the thread, the escrow and the private eval might still be worth a weekend. Attention is the asset he told developers to chase. The chain, in this telling, is no longer the excuse.
