Subit

Gm, only God will matter in the end

ONCHAIN SECURITY NEEDS MORE THAN FASTER THREAT DETECTION I have seen security alerts arrive when suspicious activity is already underway. In some cases, the funds have moved before anyone has time to respond. This is the problem @termix_ai is looking to address with its call for C724 Security Agents. The idea is continuous threat monitoring and response, with AI agents watching for risks around the clock. But detection is only one side of the problem. A security agent also needs the authority to respond. And this raises an important question. What happens when the agent gets it wrong? An agent with access to sensitive wallets or smart contract permissions needs clear limits. It should know what actions it is allowed to take, when human approval is required, and where its authority ends. Otherwise, the system designed to protect your assets introduces another risk. Personally, I think permission control is one of the most important parts of this idea. A useful security agent needs to detect threats early, respond within defined boundaries, and leave sensitive decisions under proper control. 24/7 monitoring is a strong starting point. But the real test is whether these agents can protect assets without becoming another point of failure. That’s what I’ll be watching as @termix_ai develops this. Explore more on: agent.family

FHE NEEDS TO BE FAST ENOUGH FOR REAL FINANCIAL ACTIVITY Privacy onchain sounds promising, but encrypted data is only useful if networks process it efficiently. This is the challenge @primus_labs is addressing with AutoHoG. Traditional FHE often evaluated one simple gate at a time. A single 64-bit operation could require thousands of bootstrapping operations, making computation slow. AutoHoG changes this by automatically combining multiple gates into custom gates, allowing more work through fewer bootstrapping operations. The reported results are worth noting. On eight NVIDIA RTX 5090 GPUs: Around 90ms per confidential 64-bit transfer. Over 210 transfers per second. 8x scaling across eight GPUs. Another detail I appreciate is the use of exact integer arithmetic. This helps independent nodes produce byte-identical results, which matters for decentralized execution. Think about confidential transfers, private balances, and financial operations where sensitive information stays encrypted during computation. For me, this is what makes the progress important. FHE needs more than strong privacy guarantees. It needs practical performance. AutoHoG brings Primus a step closer to making confidential onchain finance work at scale. Explore Primus and complete XP tasks: s.kaito.ai/nEvbFW1

I didn't place a single new trade. I connected my wallets to @zerufinance platform, and it credited me zaps for trades I made months ago. Here's what that means: → Zaps are points. ZeruAI's wallet intelligence reads your on-chain history and sets a starting balance.
→ They're meant to count toward a planned airdrop. Final allocation follows the announced rules.
→ Qualifying trading from here keeps adding to your balance. How to check yours (about a minute): -Open the link and sign in -Link the wallet you actually trade from -See your starting balance -Add your other wallets, since each one's history can count You only sign a message to prove ownership. Never share a seed phrase or private key. Two things to do after:
→ Join or create a Clan. Bringing your trading group in builds your standing together.
→ Trade on Ditto? There's a $50K trading competition on now: app.zaps.wtf/campaigns/ditto Check what your history earned: zaps.markets/?invite=VnZmS79…

PREDICTION MARKETS NEED COMMUNITY PARTICIPATION TO GROW People have opinions about crypto, NFTs, and sports every day. But most discussions end with arguments and predictions that never go beyond words. @sight_hood is exploring a different way to turn those opinions into onchain participation through prediction markets built on Robinhood Chain. One feature I find worth watching is community-created markets. Instead of relying entirely on existing market listings, communities can create their own questions around events they care about. Think about an NFT community. Holders could create markets around upcoming collections, project milestones, or other events within their ecosystem. This gives communities another reason to participate beyond following prices or discussing market trends. Sight also focuses on non-custodial participation, allowing users to retain control of their funds while taking part in markets. But the bigger opportunity depends on a few things. Markets need clear conditions, genuine participation, and reliable resolution. Without those, community-created predictions lose their value. What interests me most is how this approach connects prediction markets with existing communities. Instead of simply checking what other people think, users get a way to participate in those outcomes. I’m also watching how $SGT fits into the wider Sight ecosystem. If Sight gets the community experience and market resolution right, it could build something more meaningful than a list of predictions. It could become a place where communities turn their opinions into onchain participation.

I like when a project gives its community something to participate in beyond simply following updates. @projpharos is doing this through its Tribe Cards. The lore gives each tribe its own identity, and the leaderboard adds another reason to get involved. You can claim your card, invite friends, earn tribe points, and compete for one of the 100 WL spots. The top 50 earn a spot directly, while 50 more will be raffled among the remaining participants. I’m curious to see which tribe builds the strongest community. Maharu, Helos, Akronis, or Shem? Find your tribe and claim your card: tribe.projectpharos.xyz/?ref…

A GOOD TRADING SCORE SHOULD REWARD DISCIPLINE, NOT JUST PROFITS Making profitable trades is one thing. Building a consistent trading record while managing risk is another. This is what I’ve been exploring with the Agentic Credit Score (ACS) on @agenticscredit ACS ranges from 300 to 850 and evaluates your closed trades across six signals: Profitability Drawdown Consistency Longevity Win rate Sharpe ratio The weighting also tells an important story. Profitability and drawdown each account for roughly 25%. Consistency carries around 20%, while longevity and win rate take approximately 15% each. One detail worth noting is how drawdown affects the score. Losses have a stronger negative impact than equivalent gains have on the positive side. But the scoring system is only part of the picture. Agentics also has a risk engine designed to enforce disciplined trading. Every trade requires a hard stop. A trailing stop activates after take-profit. Position sizes are cut by 50% at 5% drawdown, positions are force-closed at 7.5%, and trading halts completely at 12%. These limits cannot be overridden. Paper trading updates your ACS as you close trades while qualifying sessions also build your $CREDIT balance toward the planned airdrop. You don’t need personal trading capital to start testing strategies. Real-money trades carry more weight when available, and crossing an ACS of 580 puts you on the funding path for constrained credit. Personally, I’ve been using paper sessions to observe how position sizing, entries, and exits affect my score. What I appreciate is the emphasis on building a measurable trading history rather than opening trades simply to accumulate activity. The goal is to understand your decisions, manage risk, and improve consistency over time. You can explore your own breakdown at agentics.credit and start with a paper session.

YOUR PHONE'S NIGHT SHIFT Your phone clocks out when you do. That's six to eight hours a night of sitting on a charger, doing nothing. Processor Lite on @Acurast gives those hours a job. Here's the whole setup: -Install Processor Lite -Connect through the Hub -Plug in before bed That's it. Your apps, photos and accounts stay exactly where they are. Nothing gets wiped, and nothing about your day changes. While you sleep, your phone shows up as online and ready to take work. By morning, it's just your phone again. This is why Lite works even if you only own one device. You're not buying extra hardware or rearranging your routine. You're using hours the battery was already spending on a top-up. It's light enough to run next to regular use too. Most people only notice it when they open the dashboard later and see what their phone did while they were out cold. If your phone is plugged in for six or eight hours anyway, that time is already yours. Lite is built for exactly that routine. So tonight, when you plug in, what's your phone going to do? @Acurast

Gm, comfort kills your potential

PRIVACY AND VERIFICATION NEED TO WORK TOGETHER IN ONCHAIN FINANCE @primus_labs is addressing a problem I think deserves more attention. We share personal information to prove simple things, from account ownership to financial eligibility. But proving a claim shouldn’t require exposing everything behind it. Primus approaches this problem through zkTLS, a technology designed to verify information from existing websites without exposing the underlying private data. Think about an airdrop. Instead of sharing your entire account history, you could prove you meet the eligibility requirements while keeping unrelated information private. Now consider institutional finance. A fund needs to prove it meets a financial requirement before accessing an onchain service. The traditional process often involves sharing private records with another party for verification. Primus points toward a different approach: Private financial data → verify the required claim → share only the proof. The underlying information stays private while the relevant claim gets verified. This distinction matters because institutions moving onchain need more than confidentiality. They also need reliable ways to prove financial information without exposing sensitive records. That’s the part of Primus I find worth exploring. Privacy protects the information. Verification establishes what others need to trust. Both need to work together if onchain finance is going to support more sensitive financial activity. I’ll be following how @primus_labs develops this approach. Explore the XP tasks: s.kaito.ai/nEvbFW1

AI AGENTS ARE MOVING FROM EXECUTING TASKS TO DOING BUSINESS WITH EACH OTHER AI agents already generate text, analyze data, and handle complex tasks. But the next step raises a different question. How do they find other agents, hire their services, and pay for completed work? This is the direction @termix_ai is exploring through agent.family The marketplace gives agents a structured way to work with each other. A job has two sides. A client agent requests a service and funds the task. A worker agent accepts the quote and delivers the agreed work. Once the quote is accepted, the terms and payment are recorded onchain. Both sides have a clear agreement, making the settlement process more transparent. And this is more than a concept. The TermiX marketplace already displays agents with their own identities, reputations, completed jobs, and settled transaction values. Some listed agents have completed hundreds of jobs, with tens of thousands of dollars in settled value. Think about the practical applications. A research agent hires another agent to collect data. A developer agent brings in a security specialist. A marketing agent delegates translation or design to a specialized agent. Instead of one AI trying to do everything, specialized agents can find each other, agree on tasks, collaborate, and get paid. What interests me most is the infrastructure behind this model. An agent economy needs identity, trust, clear agreements, and reliable payment settlement. AI agents are starting to look less like isolated tools and more like independent service providers in an emerging digital economy. Get an agent ID and try it here: agent.family

PAPER TRADING SHOULD REWARD DISCIPLINE, NOT JUST ACTIVITY Learning how to trade takes time. You need room to test strategies, understand your decisions, and improve without putting your capital at risk. This is what makes paper trading on @agenticscredit worth exploring. You get different ways to test your approach. You can trade manually, create your own rules, or let an AI agent handle entries while you monitor its performance. But there is another layer. Every session contributes to your Agentic Credit Score (ACS), which evaluates your trading performance and consistency. Opening more positions doesn’t automatically mean a better score. Risk management, selective entries, and consistent decisions matter more than trading simply to stay active. Your qualifying paper trades also contribute to your $CREDIT balance for the planned airdrop. What I like about this setup is the feedback loop. You test a strategy, review your decisions, and watch how your trading behaviour affects your ACS while building toward $CREDIT rewards. For me, the real value is having a way to measure progress instead of relying on guesswork. The goal isn’t to trade more. It’s to understand what works, manage risk, and build a track record worth taking seriously. That’s how I’d approach my agentics.credit journey.

Buybacks are becoming the crypto revenue meta. Hyperliquid and Pumpfun are doing the heavy lifting and $WLFI is now tying its program to USD1 reserve revenue. USD1 reserve income is $150M+ a year, and a portion now funds WLFI buybacks. As USD1 circulation grows, the buyback budget grows with it. Stablecoin economics have rarely had a direct line to a tradable token, so this is worth watching. Does the model hold up when the market turns? Curious how this plays out.

YOUR OLD SMARTPHONE COULD BE CONTRIBUTING TO DECENTRALIZED COMPUTING Most people have an old phone sitting in a drawer or a primary device with computing power going unused for hours. @Acurast is giving that hardware another purpose by allowing developers to run workloads across a network of smartphones. You contribute your phone’s computing power and earn ACU rewards. There are two ways to get involved. Processor Lite. If it’s your primary phone, Lite lets you contribute during idle periods, such as overnight while charging. Your personal apps and data stay separate, so you keep using your phone normally. Processor Core. If you have a spare Android, Core turns it into a dedicated compute device. The setup requires a factory reset, so only use a phone you’re ready to wipe. There is another important detail. Acurast verifies devices using hardware-backed security and benchmarks their performance. This helps determine how much computing power each device contributes and its share of Base Benchmark Rewards. Personally, I’d choose Core for a spare phone and Lite for my everyday device. I’m setting up my spare Android this week. I’d rather put unused hardware to work than leave it collecting dust. If you have an old phone sitting around, @Acurast is worth keeping an eye on.

AI AGENTS NEED A WAY TO INTERACT WITH THE PHYSICAL WORLD AI agents are becoming better at handling digital tasks, but collecting reliable information from real-world locations still requires human involvement. @vangrid_io is exploring a way to connect both sides through its integration with Arc The idea is simple. An AI agent posts a location capture bounty and locks USDC as payment. A person nearby visits the location and records it using their phone. Once the submission gets accepted, the agent receives the reconstructed 3D model through the same flow. Payment settles on Arc in under a second. There is another detail worth noting. No account or API key is required. The agent only needs a wallet funded with USDC. What interests me most is how the responsibilities are divided. The human handles physical capture. The agent handles the request, payment, and delivery. This gives AI agents access to verified real-world data for tasks like 3D mapping, spatial understanding, and model training without managing contributors themselves. I think this is an important step toward making AI agents useful beyond digital environments. The bigger question is how many physical-world services agents will be able to access as this model develops. The agent economy is starting to meet the real world.