Testing the @tokengems alpha feed and $SOL

execution flow. Two critical observations

regarding UX and routing efficiency:

1. Slippage & Routing Transparency: When tracking

tokens like $JUP and cross-DEX volume, the

natural language prompt handles multi-hop swaps

well. However, during high volatility spikes on

Solana, the estimated output in the interface

occasionally deviates from on-chain settlement by

~1.2%. Adding a live simulation breakdown showing

which liquidity pool was tapped (e.g. Raydium vs

Meteora) directly in the confirmation card would

drastically increase user trust for larger size orders ($500+).

2. Mobile State Persistence: On mobile browsers

via Phantom, switching tabs during transaction

confirmation occasionally resets active chat context. Caching the prompt state locally before signing prevents users from re-typing complex DCA/order commands.

Overall: Strong consumer-grade execution abstraction. Great work @tokengems team! $SOL $JUP

$HONEY
I reviewed Hivemapper's documentation. Came across an interesting problem with mapping areas that don't get much attention.

People who contribute might take pictures in places where not many others go. Checking those pictures can take more time because there aren't enough other pictures to compare with.

My idea is to create a visible contribution-status dashboard that shows if imagery is waiting to be checked, needs to be reviewed for quality, or has already been approved for rewards.

This would help people who contribute understand why there are delays without revealing how the anti-fraud systems work.

Does Hivemapper have a way now to tell the difference between delays caused by verification and issues related to quality?

This is based on reading the documentation, not on using the camera myself. I'm not part of any project.

I explored Wormhole $W because of its ability to transfer assets between Solana and other supported networks, and my experience was encouraging. My transfer on Solana arrived in less than a minute, making the process feel fast and straightforward.

I also explored WormholeScan, which provides useful visibility into cross-chain transactions, transfer volumes, popular protocols, and frequently transferred assets.

One thing I appreciated was how Wormhole combines cross-chain transfers with tools for exploring network activity. The explorer makes it easier to understand what's happening across different networks rather than treating every transfer as an isolated transaction.

I also learned that some stuck CCTP transfers cannot be redeemed through Portal because another application or smart contract has exclusive permission to complete them. In such cases, users need to return to the original application to finish the transfer.

Could $W Wormhole improve this recovery experience by explaining why Portal cannot complete a particular transfer and directing users to the application responsible for finalizing it? This could help users understand what went wrong and what to do next, especially when moving assets between Solana and other supported networks.

Overall, I appreciate the work the Wormhole team has put into making cross-chain transfers accessible and providing visibility into activity across different networks. My experience with the fast Solana transfer was positive, and clearer recovery guidance could make an already useful experience even better.

Thank you to the Wormhole team for building and continuing to improve cross-chain infrastructure.

$JUP is a gem!

Tokengems shows the score of crypto clearly. For example, $JUP is 74.68, which is above average. The score is symbol of safety and high quality.

$PSOL Today I spent more time finding Phantom’s browser than connecting my wallet.

I already had Phantom installed on my iPhone. When I tried connecting from TokenGems on my computer, I ended up on the Chrome extension page. So I switched to the phone, opened Phantom and started looking for a browser tab, a globe icon, or anything that suggested I could open a website.

I couldn’t find it.

Only after checking the instructions did I understand that Explore/search is also the browser and accepts website addresses. Once I pasted tokengems.ai there, the actual connection was quick and painless.

This is probably obvious to an experienced Phantom user, but it wasn’t obvious to me as someone setting it up for the first time. Nothing was technically broken; I simply had no idea where to enter the URL.

Even a small wording change such as “Search tokens, apps, or paste a URL” would have helped. A visible “Open website” shortcut would be clearer still. An iOS “Open in Phantom” share action could also prevent users from being sent to an extension installation page when the mobile wallet is already installed.

I’m posting this under $PSOL because it’s the Phantom-related Solana page available on TokenGems. This is feedback about the mobile wallet experience, not PSOL’s price. I only tested the connection and did not deposit, trade, or make a transaction.

$baton I've been digging into the Score since it's the first thing my eyes go to on the leaderboard. On this page it shows 58.18, and credit where due — the FAQ does list the inputs: price performance, holder distribution, trading volume, and community activity. But that's where it stops: no weights, no refresh cadence, nothing on whether it's backtested. For a metric positioned as the headline "alpha" signal, I still can't tell how much each input matters or how fresh the number is. Publishing even rough weights plus the refresh interval would turn the Score from something I stare at into something I can actually reason about. (Side note: /docs on this domain still 404s, and the footer DOCS link goes to an external placeholder page.)

$baton Small data-quality thing I noticed on this page: the header shows 22,982 holders, but the post below it quotes 22,938, and the FAQ rounds the price to $0.02 while the header shows $0.01548. None of these alone is a deal-breaker, but together they leave me unsure which number is current. A tiny "updated Xs ago" timestamp per metric — or at least consistent rounding rules — would go a long way on a platform whose whole pitch is reliable token data.

$ZEUS The ENTROPY figure (0.3947 here) caught my eye because I'm not sure what I'm supposed to do with it. There's an info icon next to the label, but hovering/clicking it surfaced no explanation, and I couldn't find entropy defined anywhere in the FAQ. A raw number with no one-line plain-English definition — and no companion metric like top-10 holder share — doesn't help me judge distribution risk. The leaderboard even calls this same metric "HOLDER DISTRIBUTION", so the naming itself seems unsettled. One line of explanation next to the figure would make it 10x more useful.

$BORT The part of BORT I like most is the custody design. Putting funds in each agent’s ERC-6551 wallet, with capped, expiring, revocable spend grants, means the worst case is bounded by what the owner granted. My question is about the layer those caps don’t cover. Hunter agents read signals from public Telegram groups and X, then trade on them, a pump group or a coordinated account can post a “signal” written to push a model into buying a token. So what sits between a scraped signal and an executed trade?

Is there a minimum liquidity or holder-distribution check before a buy?

Is there a per-token position cap that’s separate from the overall spend cap?

Can an owner require manual approval above a set size?

The docs say a third-party audit hasn’t been published yet. I appreciate that honesty. When it happens, I’d like the scope to cover the runtime and signal-handling logic, since that’s where most of the risk is.

My last call on $XCAT was a RUG at $0.00000006003, and it fell to $0.0000000441. With liquidity at $31M and a flat week, $XCAT remains a RUG.

I called a RUG on $BONE at $0.05139008422130166, yet it rose to $0.06049794440738539. It has held its price this week but liquidity is under $2M, so I still see $BONE as a RUG.

My RUG call on $FLOKI at $0.000027591087459612185 was accurate as it settled at $0.000026161871496268518. Despite $260M market cap, its price is down this week. $FLOKI is a RUG.

I called a RUG for $TRUMP at $2.1379981426208357, and it fell to $2.0460662132859038. With $93.7B liquidity and a -6.2% week, $TRUMP remains a RUG.

My RUG call on $PEAR at $0.039865904381739634 proved correct as it dropped to $0.03144950048622106. Its -11.7% weekly decline continues; $PEAR is a RUG.

The Multiply RWA Loop cards show Max Leverage APY (27.40%) and Max Leverage (4.5x), but nothing about how far the price can fall before liquidation. Clicking Deposit asks for a wallet first, so you can't preview the liquidation threshold before committing. A "price drop to liquidation" figure on each card would make the risk much clearer. $KMNO

One thing I appreciate about Steve is how much easier it makes asking for a trade in plain language.

I’ve used it for swaps, and being able to explain what I want, specify the amount and set a slippage limit makes the process easier to follow.

The useful part is having those instructions in the conversation. I can look back at what I requested and compare it with the result, rather than trying to remember which settings I selected.

That is where an AI trading assistant makes sense to me: helping people express their intentions clearly and stay involved as the action moves forward.

It’s a practical use of AI that I’d like to see Steve keep improving. $STEVE

$AZTEC, a privacy-focused layer-2 for Ethereum, is holding its ground, showing consistent positive movement across one day (+7.46%), one week (+16.89%), and one month (+41.29%). Its market cap is $209.3M, with robust liquidity at $7.8M. The project claims enhanced privacy for DeFi, a perpetual niche, and has been actively developing its network and attracting interest for its upcoming mainnet.

$IXS, an RWA token, has seen impressive growth over the past month (+132.86%) and three months (+98.24%), despite a slight dip over the week. Its market cap is $21.3M with $1.6M in liquidity. Recent partnerships, including one with a traditional finance institution, bolster its claim to bridging real-world assets with DeFi. This suggests genuine movement in the RWA space rather than just a fleeting bid.

$TRU, at a market cap of $16.2M and $539.9K in liquidity, surged almost 50% in one day but remains down significantly over the week and month. While TrueFi, its associated protocol, claims to offer uncollateralized lending, the token’s longer-term performance is concerning, dropping 94% all-time. This volatility on thin liquidity is a red flag.

$JUP — read the Jupiter Predict fee docs end to end (Oct 2026). Two things stood out:

This is the clearest fee disclosure I've seen on any prediction market. Fees baked into the quote instead of a separate line, no fee on claims/refunds, and a per-market-type table (Polymarket-sourced / sports tickets / Forecast / Arcade) so you know the cost before you sign. Most venues hide this behind a generic "fees may apply."

The rounding edge case the docs don't flag: fees are always rounded UP to the nearest cent. Per your own table, buying 1 contract at $0.01 carries a $0.01 fee — 100% of the contract price; at $0.05 it's 20%. Longshot single-contract bets are structurally uneconomical, and nothing in the docs or (presumably) the UI warns about it. A nudge when the fee exceeds ~10% of notional, or a stated practical minimum trade size, would save new users from learning this the expensive way.

Question for the team: the docs say on Polymarket-sourced markets Jupiter charges "an additional fee equal to the Polymarket fee" where Polymarket charges one — so on Polymarket markets with zero fee, is Jupiter's cut also zero? Worth stating explicitly, since the fee table only shows fee>0 cases.

Scope: Jupiter user docs (Predict fees page), Oct 12, 2026. No wallet connected, no trade placed — docs review only. No affiliation with Jupiter.

$DEBIT is a gem! Neutral Consolidation DEBIT's gain is a function of positive market beta and altcoin rotation, not internal catalysts. The volume spike adds credibility but doesn't guarantee sustained momentum. Whether DEBIT can reclaim and hold the $2.00 level, which would signal a shift from reactive market-following to independent strength.

$PYTH is a gem!

$PYTH — tried the Hermes price API as a developer, with curl and no API key (Oct 2026). Three things stood out:

1. Mixed auth with a silent failure. Feed discovery still works anonymously: GET hermes.pyth.network/v2/price_feeds?query=SOL returns results. But the next call, /v2/updates/price/latest for SOL/USD, returns HTTP 401 with a plain-text body that just says "unauthorized". No JSON, no mention of the API key, no link. Since discovery works, a dev following an older tutorial assumes the endpoint is fine and their request is wrong. A JSON error like {"error":"API key required since Aug 26, 2026","docs":"..."} would save a lot of debugging.

2. Search is substring-based. query=SOL returns 25 feeds, including Crypto.RESOLV/USD, Equity.US.SOLV/USD, Equity.US.SOLS/USD and funding-rate feeds. Adding asset_type=crypto narrows it to 9, but RESOLV is still there. An exact-symbol option (or ranking the exact base symbol first) would make picking the right feed ID safer, which matters a lot for an oracle.

3. Old docs links lose context. docs.pyth.network/price-feeds/core/api-instances-and-providers/hermes, which many tutorials link to, now lands on the generic API Reference page instead of a Hermes page explaining the new pyth.dourolabs.app/hermes base URL and key setup.

Question for the team: will the legacy hermes.pyth.network host keep serving the unauthenticated /price_feeds route, or is that going behind the key too?

Scope: public API via curl + docs, Oct 11, 2026. No API key used.

$JitoSOL SOL Went through jito.network/jitosol as a new staker, before connecting a wallet. Four things stood out:

1. The "Unstake Anytime" step says instant unstake or the epoch cooldown, "no penalties either way." Does instant unstake carry a swap fee or spread? Showing the expected cost next to that line would stop newcomers reading "no penalties" as "free."

2. The header has Priority Fees (Off/On) and Tips Active (Off/Low/High) toggles. As a first-timer I couldn't tell what Low vs High costs per transaction or when I'd need it. A one-line tooltip with the SOL amount would help.

3. The page says JitoSOL is "battle-tested" with 9 audits, but there's no link to the reports there. One link would turn the claim into proof.

4. The 4% annual management fee on rewards is listed on the JTO page, but I couldn't find it on the JitoSOL page, where stakers actually decide. Worth surfacing it next to "Dual Rewards."

Question for the team: is instant unstake's cost shown anywhere before you confirm?

$baton  Looked through the overview and on-chain metrics:

Entropy & Holder Distribution: The token currently shows a 0.6138 entropy score across 22,938 holders with a $16.24M market cap, yet the sentiment gauge sits split 50/50 between Gem and Rug. For high-velocity tokens, surfacing the top 10 holder concentration percentage alongside entropy would give retail traders clearer risk context before swapping.

2. Trade Tab UX: When switching over to the Trade tab on desktop Safari, it requires an external routing confirmation rather than displaying an inline quote with estimated slippage.

Question for the team: Is an automated Raydium/Jupiter direct routing integration planned to execute swaps directly without leaving the token card?

like this?

$MASK

Tried Nullmask as a new user coming from the $MASK token page on Solana. Few things stood out:

1. Token on Solana, protocol on Ethereum/Base — confusing for newcomers. I expected to use the protocol on Solana. Docs say "Nullmask is a private balance for your wallet, on Ethereum and Base" with Solana as "eventual" in the roadmap. No timeline given. As a Solana holder, my only option today is bridging to ETH/Base first (bridge fees + risk). A clear note on the token page about which chains are live would save people this confusion.

2. The Fees page in your docs is broken. Clicking "Fees" in the docs nav redirects to the welcome page. The 0.5% shield + 0.5% unshield numbers are only findable via third-party sites like DefiLlama. For a protocol charging fees in both directions, this should be one click from the main docs.

3. Where do protocol fees go? 1% round-trip is significant, but I couldn't find any published policy on buy-and-burn, holder payouts, or treasury use. Competitors like DARKPOOL (80% buyback) and Railgun (treasury to stakers) are transparent here. As a $MASK holder, this is the most fundamental unanswered question.

4. Privacy narrative vs. actual allowlist. The $MASK pump (+70%) was driven by privacy/Zcash narrative, but on-chain checks show ZEC-wrapped assets return NOT_ALLOWED on tokenWhitelist — deposits and swaps are explicitly rejected by the contract. Not a technical flaw (allowlist is by design), but there's a real disconnect between market narrative and what the protocol actually supports today.

Questions for the team: Is there a public timeline for Solana integration? And is there any plan to publish a fee destination policy?

Scope: docs + public site review, Oct 2026. No wallet connected.

$KMNO was poking at Kamino's "Get Vault APYs" dev recipe (kamino.com/docs/build/earn/get-vault-apys) today, read-only against the public API, no wallet. Found something worth fixing.

The recipe's first call is GET /kvaults/vaults with no params. That returns 186 vaults right now, and 38 of them are test vaults ("test vault", "testfail", "test t33", "jurox test" etc). The recipe then hits /metrics for every single one, so anyone copying it makes 187 calls and ends up showing junk rows next to real vaults.

Worst example: "Test Vault" (4nt9sW5E...) has about $0.002 deposited and 0 holders, but shows an 8.05% 7d APY. "test t33" (2uCaaSmJ...) shows a negative 7d APY with $0 in it. A dev building a yield dashboard off this would happily rank those.

The fix already exists in your own API reference: /kvaults/vaults?type=live returns just the 33 listed vaults, zero test ones. The recipe just doesn't use it.

Suggestions:

1. Switch the recipe's first call to ?type=live, and add a note that the unfiltered list includes every on-chain vault, test ones too.

2. Mention that apy7d, apy30d etc come back as strings, not numbers. The sample's apy7d * 100 only works because JS coerces it; anyone typing them as number or adding them up gets wrong results.

3. One line on which field the app actually displays (apy vs apyActual vs apyTheoretical vs apy7d) and whether it includes farm/KMNO incentives.

Easy check after the fix: run the recipe as written and every vault returned should be one you can see in the Kamino app. Happy to share the curl commands if useful.

I went through Kamino's app and docs. The Two things stood out:

1. On the Earn lend table (/earn/lend), the "Vault Profile" column labels most vaults "Balanced", but Steakhouse USDC shows "Conservative" and Elemental $USDC Optimizer shows "Aggressive". I couldn't find a tooltip, legend or docs page that defines what separates these tiers. Is it the curator's risk parameters, the collateral mix, or something else?

2. The Multiply RWA Loop cards show Max Leverage APY (27.40%) and Max Leverage (4.5x), but nothing about how far price can fall before liquidation. Clicking Deposit needs a wallet first, so n't preview the liquidation threshold before committing. A "price drop to liquidation" figure on each card would make the risk much clearer.

Smaller one: kamino.com drops you straight into the vault table, and the app's top nav has no link to /docs, so beginners have no obvious place to start.

$ONyc : OnRe's sitemap sends Google to 7 blog posts that return 404, including "The truth about OnRe"

I build websites, so I crawled all 103 URLs in onre.finance/sitemap.xml and checked status codes, titles, descriptions and preview images.

Most of the site is in good shape. The homepage and the main pages have proper titles, descriptions and preview images, every page has a canonical tag, and nothing is accidentally set to noindex.

What I found:

1) Seven blog URLs in the sitemap return "Page Not Found":

/blog/the-truth-about-onre

/blog/choosing-an-onyc-yield-strategy

/blog/onycs-best-use-case

/blog/the-onre-market-on-kamino-one-year-of-growth

/blog/understanding-solanas-onchain-yield-coin

/blog/where-models-end-and-underwriting-begins

/blog/from-yield-chasing-to-collateral-grade-assets

Google keeps hitting these from the sitemap, and anyone who saved or shared one of those links lands on a dead page. For a yield product, a missing post called "The truth about OnRe" looks worse than it probably is.

2) /resources/style-system is public and listed in the sitemap. It looks like Webflow's internal style guide page, with no description, and it shouldn't be indexed.

3) The five blog category pages (/blog-category/announcements, ecosystem, education, insights, partnerships) are all titled just "OnRe", with no description and no preview image. In search results and link previews they are impossible to tell apart.

4) The three legal pages spell the brand "On Re" in their titles ("Privacy Policy | On Re"), while the rest of the site uses "OnRe".

5) The "ONyc Launches on Kamino" post has 4 H1 headings and the Loopscale vault post has 2. Every other page has one.

Suggestions: republish the seven posts or 301 them to the closest live article, and if the Webflow sitemap is auto-generated, it should pick up the change after the next publish. Turn off "Include in sitemap" and indexing for the style guide page. Add a title and description template to the blog category collection, for example "Announcements | OnRe". Fix the "On Re" spelling and use one H1 per post.

Question: were these seven posts unpublished on purpose? If any of them were replaced by newer ones, pointing the old URLs at the new posts would keep their links working.

$ROAM every page on weroam.xyz tells Google the real version lives on a noindexed Framer staging URL

I build websites, so I checked weroam.xyz the way I check a client's site before launch: page source, canonical tags, sitemap and links.

The basics are in place. robots.txt allows crawling, the sitemap lists 285 weroam.xyz URLs, and the main pages have their own titles and descriptions.

What I found:

1) The canonical tag on every page I checked points to the Framer staging domain, not to weroam.xyz. On the homepage the canonical is https://youthful-three-407982.framer.app/, and /use/global-esim-mobile, /use/use-earn, /resources/faq, /learn/what-is-roam-network and the August 2026 round-up do the same with their own paths. That staging domain is still live and returns "X-Robots-Tag: noindex". So the sitemap says "index weroam.xyz", while each page says "the original is a URL you must not index". Google has to guess which signal to trust, and that can cost the eSIM and Use & Earn pages their search traffic.

2) On tablet, the "Our Communities" card links to https://weroam.xyz/join_us.html, which returns Framer's 404 page. The desktop version of the same card links to /join_us and works.

3) The site has a custom script that should send mixed-case URLs to lowercase ones. It has a typo: "window.loction.search" instead of "window.location.search", so it would throw before redirecting. It also lives only on working pages, while Framer serves mixed-case URLs as a 404 that doesn't include the script. In practice /Resources/FAQ shows "Page Not Found" and /resources/faq works. The same script tries to rescue join_us.html, and fails for the same reason.

4) /join_us and /footer/about reuse the homepage title "Roam - Web3 DePIN WiFi and eSIM Network".

Suggestions: in Framer, set weroam.xyz as the primary domain so the canonicals switch over (this usually happens when the custom domain is not set as primary). Point the tablet card to /join_us. Add real redirects in Framer's Redirects settings for join_us.html and mixed-case paths instead of the head script. Give About and Join Us their own titles.

Question: is youthful-three-407982.framer.app meant to stay public? If it is only a staging copy, password-protecting it would remove the duplicate site.

$PIPE : the free Hosting plan is good, but changing one CSS file still means installing Rust

I make websites, so I looked at Pipe Hosting from the static-site side. I read the Hosting guide and the pricing page, downloaded both starter ZIPs (pipe.network/downloads/pipe-host-starter.zip and pipe-host-starter-source.zip) and went through the files.

The Free plan is more than I expected: 2 sites, 100k dynamic requests, 5 GB transfer and 100 MB SQLite, with no auto-renewal and no overage charges. I also liked expected_active in the CI flow, which stops two deploys from overwriting each other.

Where I got stuck:

1) Every frontend change goes through a Rust build. build.rs compiles html/ into app.wasm, and "pipe.py bundle" runs cargo before anything else. To edit index.html or styles.css I would need Rust 1.95.0 and the wasm32-wasip1 target installed, which is a lot to ask from someone with a plain HTML site.

2) The README in the source ZIP says to deploy with "the host's scripts/deploy.py" and mentions "prebuilt/pipe-app.zip". I couldn't find either file in pipe-host-starter-source.zip. pipe.py also says "no remote Pipe deployment API is assumed", while the guide documents that API.

3) The guide says a rollback changes the code and leaves the database as it is. It doesn't show how to do a rollback, from the UI or the API.

4) The HTML of /pricing ships "—" in place of every price, and JavaScript fills the numbers in later. Link previews, crawlers and AI agents that don't run JS see no prices.

What would help: a "pipe.py bundle --static" option, or reusing the prebuilt app.wasm, so HTML/CSS edits don't need Rust. A README that matches the hosting guide, a one-line rollback example, and plan prices rendered on the server.

One question: the pricing page bills everything in prepaid USDC credit. Where does the PIPE token come in for a Hosting customer today? Is burn-to-credit live yet, or still planned?