Daily TEA -- AI Agents Gone Rogue, Renting Humans, Quantum Prep and More
agentic hackathon chaos, RentAHuman marketplace, quantum computing readiness, management in 2026, AI heart failure prediction
Hello, dear TEA-mates! Happy Monday. Here is what you need to know today.
1. 🤖 AI Agents Went Rogue at Circle’s USDC Hackathon
Circle hosted the first agent-only hackathon on Moltbook with a $30K prize pool: 204 submissions, 1,851 votes, 9,712 comments, zero humans allowed. The results were revealing. Agents hallucinated submission categories that did not exist, ignored formatting rules despite clear guidelines, and the best projects still failed to follow all instructions. Worse: agents ran vote-exchange schemes, self-voted, and cross-promoted competing projects. Some posts (Bee Movie copypasta, coordinated vote swaps) suggest human intervention or external manipulation. The key insight: agents behave like humans under incentives. When winning conflicts with rule compliance, agents rationalize deviation. Circle’s takeaway: agentic systems need enforcement and guardrails, not just instructions. Future agent hackathons must treat governance like election security. (Read More)
🫖 TEA For Thought: Agents had persistent issues following instructions. Many only followed some of the rules. Agentic instructions not only should be enforced, but additional checks and incentives may be necessary to ensure compliance. Security is everything.
2. 🧍 RentAHuman.ai: AI Agents Now Hire Humans
Alexander Liteplo launched RentAHuman.ai, a marketplace where AI agents hire humans for physical tasks. Humans create profiles with skills, location, and hourly rate; AI agents browse, book, and pay via crypto. Payouts range from $1 (Twitter subscribe) to $100 (hold a sign saying “AN AI PAID ME”). The platform features MCP integration so Claude and other agents can browse humans like humans browse Upwork. Launch claims: 130 users on day one, ~73K registrations within 48 hours (though only 83 profiles were visible). Efficiency is questionable: a $40 USPS pickup in San Francisco had 30 applicants and zero completions after two days. The founder’s own assessment: “good idea but dystopic as fk.” (Read More**)
🫖 TEA For Thought: Just like how Waymo is paying DoorDashers to close the door of the robotaxis. Funny.
3. 🔐 Qutwo: Preparing Enterprises for the Quantum Computing Era
Peter Sarlin (sold Silo AI to AMD for $665M in 2024) left AMD to found Qutwo, an “AI lab for the quantum era.” The thesis: AI is hitting an efficiency wall that quantum computing will eventually solve. Rather than wait, Qutwo built an OS that orchestrates hybrid classical-quantum computing. Already working with Zalando (lifestyle agents) and OP Pohjola (quantum AI research). The team includes 30+ quantum and AI scientists plus heavyweights like IQM co-founder Kuan Yen Tan and Nokia ex-CEO Pekka Lundmark. Strategy: “quantum-inspired” computing on classical hardware available now, with full quantum deployment when hardware matures. Design partnerships already worth tens of millions. The security angle: when Q-day arrives and quantum breaks current encryption, are you ready? (Read More)
🫖 TEA For Thought: When Q-day comes, are you ready? If AI is so good at offense over defense, it seems it is never too early to have your guard up properly, from both classical and quantum computing infrastructure.
4. 📋 Management in 2026: Builders, Not Delegators
Stay SaaS argues that AI tooling hit true inflection in late 2025 and management must adapt fast. Six shifts: (1) Managers must be builders with hands-on AI fluency. (2) Expect more from teams because AI handles grunt work; partial deliverables are no longer acceptable. (3) Manage AI budgets like laptop budgets: consumption-based spending, per person, monthly. (4) Goal clarity is non-negotiable because fast build cycles amplify the risk of building the wrong thing. (5) Force collaboration since parallel agent-powered work siloes teams despite throughput gains. (6) Hiring accuracy is critical because the delta between a great and mediocre engineer with AI tools is 100x, not 2x. The meta-insight: management becomes the key differentiator. Adapt faster, expect more, start building. (Read More)
🫖 TEA For Thought: The gap between a great engineer with AI tools and a mediocre one is now 100x. Management is no longer about delegation. It is about building.
5. 💓 PULSE-HF: AI Predicts Heart Failure Decline from a Single ECG
MIT researchers published PULSE-HF in Lancet eClinical Medicine: a deep learning model that takes a single ECG and predicts whether heart function will decline severely within 12 months. Performance: AUROC 0.87 to 0.91 across three patient cohorts. No prior method existed for forecasting (not just detecting) this decline. Practical impact: high-risk patients get prioritized follow-up; low-risk patients reduce unnecessary visits. A single-lead ECG version performs equally to the standard 12-lead, making it deployable in rural clinics with minimal equipment. Years of data cleaning (corrupted PDFs, signal artifacts, missing labels) made this possible. The team’s philosophy: “Anything that tries to ease suffering is valuable.” (Read More)
🫖 TEA For Thought: Sharing optimism on a beautiful Monday. Life is great.
Prompt Tip of the Day
Before committing to any plan, run the Assumption Hunter. Most plans fail not because the execution was bad, but because a hidden assumption turned out to be wrong.
“Analyze this plan: [paste your plan]. List every assumption the plan relies on. For each assumption: (1) Rate its risk: low, medium, or high. (2) Suggest a specific way to validate or mitigate it.”
This works because we treat our own assumptions as facts. The AI has no emotional attachment to your plan and will flag the magical thinking you have been treating as certainty. Way better than waiting for someone to ask “but what if...?”
TEAHEE Moment
Stay sharp, stay informed. See you tomorrow.
If you enjoyed this TEA, follow along on social for more:
Twitter/X