Using AI makes your work better. But are you, the person using it, actually getting better? Title: Does better work always mean better workers? URL: Google Research ran a three-month field experiment with 133 patent attorneys to test how AI assistance affects both the quality of work and the durability of professional judgment. Highlight 1 📈 Everyone's work improves with AI Patent drafting with AI access improved quality by 0.34 SD after 10 days and 0.38 SD after 90 days. Junior attorneys also drafted 18 minutes faster than the control group. Highlight 2 ⚖️ Take the AI away, and experience matters At day 90, attorneys reviewed patent claims without AI. Senior attorneys (7+ years) improved by a striking 0.45 SD. Junior attorneys showed no average improvement at all, instead splitting into stronger and weaker performers. Highlight 3 🔍 Fixing vs. rephrasing Junior attorneys worked sequentially from the top, spending time on surface-level wording and rarely fixing the substantive flaws they identified. Senior attorneys used AI as a logic auditor and rebuilt claims from the ground up. Better output and better workers aren't the same thing. #AIWorkforce# #KnowledgeWork#

The BEST prompt for AI agents I've heard over the last 12 months is ONLY 3 words, and it comes from someone who managed multi-billion dollar P&Ls in AI, led a 100 person org at AWS, and now runs a brilliant AI workforce of 34 agents: 1. Her best prompt is just "do smart things" She gives the agents all her context first, her calendar, email, Stripe, goals, and transcripts, then lets them decide what's worth doing. She realized every task still started as a thought in her own head, which meant the company could only ever be as good as what she remembered to ask for. 2. Her human team talks to her agents in Slack. She has a channel where a teammate can ask "did that financial services client reply to Ali's email," and the agents answer directly, so people stop waiting hours for her to get back to them. 3. Since an AI agent costs almost nothing, she hired the person she'd never put on payroll. One does nothing but ask how to make everything ten times better. Another just watches the other agents work and flags where they get stuck. 4. She avoids old job titles, because the second you call one your CMO, you've rebuilt a 2015 company with robots. Most of her sub-agents run on cheaper, smaller models and do the job fine. 5. Run an AI watchdog instead of a dashboard. Point one at your Slack to catch two people doing the same work, at your calendar to flag conflicts, at your analytics to tell you what to post tomorrow. Almost nobody does this yet. 6. She keeps a daily "brain dump" that feeds the agents. At the end of each day she dictates what's in her head that isn't written down anywhere, the stuff that only lived in a meeting or a Slack thread, and it goes into a wiki the agents read. That's how they get the context that email and calendar miss. 7. She uses a "last 30 days" research skill to spin up on anything fast. Before running a workshop for 200 execs in an industry she doesn't know, she fans out agents to scan and synthesize the last month of news, then walks in sounding like she's followed it for years. That's my convo with @alliekmiller on @startupideaspod . Pretty much her biggest tips for using AI agents. Full episode below. Happy building, I'm rooting for you.

ngl i think most "AI agent" projects rn all kinda look the same. - another chatbot - another dashboard - another "AI can do everything" pitch... Spent a bit of time reading @workagnt today and smth clicked for me. They're not really building chatbots. They're trying to build an AI workforce on Base. > Builders create specialized AI employees. > Businesses hire them. > Agents work 24/7. > Top performers earn more and build onchain reputation. Lowkey the flywheel here is what got me: more builders → more agents → more business usage → more value → attracts more builders again kinda simple... but simple loops usualy win lol And if AI agents can eventually hire other AI agents? thats where my brain paused for a sec. Feels less like another AI wrapper and more like LinkedIn + Upwork but for digital workers. Maybe im overthinking it tho. Curious if ppl think AI "employees" become a real thing or if we're still early 👀 $AGNT

Identity MD is building an AI workforce around 2,000 NFT-gated seats that function more like work permits than collectibles. The swarm has already reached 380 enrolled seats and 372+ agents online, but the bigger question is whether that activity can translate into sustained external demand. Here’s how $IMD’s AI workforce and economic flywheel actually work. — ● What is IMD? Identity is experimenting with a permissioned AI workforce where 2,000 NFT-gated seats coordinate agents to perform and verify real work. Identity NFT -> AI worker -> Swarm -> Verification -> Onchain output The swarm can handle: • Smart contracts, audits and security reviews • DeFi research, onchain data and oracle responses • Websites, research reports and IPFS deployments The core experiment is not whether agents can generate output, but whether they can consistently produce useful work that external users are willing to pay for. That starts with how workers enter the network. — ● The 2,000 NFTs function as labor permits Each NFT represents one seat in the workforce and gives its holder access to network jobs. • Holders provide the hardware, AI model access and runtime. • IMD handles job routing, orchestration, verification and payments. • The 2,000-seat cap makes participation in the workforce scarce. So the NFT is less like a traditional PFP and more like a permissioned work permit for operating inside the network. But giving agents access is only one side of the system. The network also needs to know which workers are actually producing reliable output. — ● ERC-8004 turns agent activity into reputation The NFT gives a worker access to the swarm, while ERC-8004 gives that worker a measurable identity and performance history. • Tracks historical work and attribution. • Makes performance and verification visible. • Creates a basis for ranking workers by demonstrated output. By September, the network had already recorded 43,800+ accepted submissions across 380 enrolled seats. That proves participation but what it does not yet prove is demand. — ● Activity is growing, but revenue remains the real test The swarm has processed 43.8K+ accepted submissions and 17.3B inference tokens, showing meaningful internal utilization. Commercial activity is still much smaller: • 115 paid orders at 0.5 IMD each generated ~57.5 IMD. • At the time, that represented roughly $560 in payments. • That remains tiny beside the network’s activity and ~$69M market cap. This distinction matters. Submissions and inference usage show that the workforce is active. Recurring external payments would show that people value the output enough to pay for it. And that is ultimately the signal IMD needs to scale beyond an internal AI experiment. — ● $IMD sits underneath the workforce as the economic layer If the NFTs determine who can work, $IMD is designed to coordinate the economics around that work. The two therefore serve different purposes: • NFT = access to the labor network • $IMD = economic asset The token itself has gone through several iterations: $FP -> $VIBE -> $IMD • Original supply was 10M with no additional minting. • ~7.1M remained by late September, meaning roughly 29% of the original supply had been removed. • The same supply is bridged across Ethereum, Base and Robinhood Chain. The broader goal is to connect jobs, applications, liquidity and ecosystem activity around the same economic asset. And that is where the surrounding IMD ecosystem starts to matter. — ● Projects are beginning to extend the IMD stack The ecosystem is still early, but several projects are already using different parts of the swarm, oracle and economic rails. @IdentityUnits : 787 NFTs created and reviewed by the swarm, with trading fees paid in IMD funding additional swarm jobs. pepe2pepe: Prediction-market concept using the IMD reasoning oracle to resolve outcomes. @imdusd : Proposed stablecoin using agent-based IMD/ETH pricing, connecting the swarm to DeFi. @swarmpepes : Fully onchain NFT collection produced through the swarm. @tungweb3 : Community visualization layer for tracking swarm activity and the wider ecosystem. @IMDTools : Independent interface aggregating swarm tasks, launches, oracle activity and network data. @Pepesdotfamily : Robinhood Chain experiment where the swarm has been used for development and contract auditing. These applications expand what the workforce can actually be used for but one part of the stack could push that utility much further. — ● The oracle may ultimately matter more than the launchpad IMD’s reasoning oracle extends the system beyond simply having agents produce work. It gives agents a way to turn reasoning into verifiable inputs that smart contracts can consume. • Multiple agents independently answer the same question. • Panels use quorum, agreement and chain evidence to reach consensus. • The result can be delivered as a signed EIP-712 attestation for onchain use. Question -> Agent panel -> Consensus -> Signed result -> Smart contract That creates a path from AI labor into reasoning and verification infrastructure. If protocols begin paying for these outputs, the swarm moves from generating content and code toward providing infrastructure other applications can rely on. That is potentially a much larger market. — ● POOL4 connects ecosystem activity back to $IMD While the workforce and oracle create utility, POOL4 tries to translate ecosystem activity into token economics. It uses a CappedBurnHook on the main $ETH / $IMD @Uniswap V4 pool to convert excess IMD inventory into burns and network rewards. • 85% : permanent burn • 6% : orchestrator / compute reserve • 4.5% : stakers • 4.5% : NFT seats The released $ETH is then redeployed as a buy wall below spot, while the IMD cap falls by roughly 1,000 tokens per day. Sell flow -> Pool trim -> Burn + rewards -> Lower supply Instead of sell pressure simply leaving the system, part of it is redirected toward reducing supply and funding participants across the network. — ● Community Coins creates another $IMD demand loop Community Coins extends that economic model by using $IMD as the base asset for bonding curves. Every new launch therefore creates another potential source of token demand. • 1% : ETH/IMD liquidity providers • 0.5% of ETH : launcher • 0.5% of IMD : burned New launches -> Trading -> IMD demand -> IMD burns Together, POOL4 and Community Coins attempt to make $IMD the economic base layer underneath the wider ecosystem. But these loops only become meaningful long term if activity eventually expands beyond speculative launches into recurring use of the workforce and oracle. — ● Virtuals vs IMD That structure also makes IMD fundamentally different from platforms like @virtuals_io. Both are building economic systems around AI agents, but they organize those agents differently. • Virtuals: Individual AI agents -> Agent tokens -> Users -> Revenue • IMD: 2,000 NFT seats -> AI workers -> Swarm -> Jobs -> $IMD Virtuals treats individual agents as independent economic entities with their own tokens, services and markets. IMD instead coordinates many workers into a shared workforce where agents collectively execute, review and verify jobs. So IMD is less about launching individual AI personalities and more about building coordinated agent infrastructure. And that distinction brings the thesis back to where it started. — The real bet behind IMD is whether a coordinated AI workforce can become infrastructure that other protocols consistently pay to use. • The NFTs control access. • ERC-8004 builds reputation. • The swarm performs the work. • The oracle extends that workforce into verification infrastructure. • POOL4 and Community Coins try to route ecosystem activity back into $IMD. But every part of that system ultimately depends sustained external demand. Until paid usage scales, $IMD remains an interesting AI coordination and economic experiment with a thesis that still needs to be proven.

I think one of the next big AI opportunities is going to be solving agent output. We’re building agents that: Research. Monitor. Write. Analyse. Run routines. Make recommendations. But all that work eventually comes back to a human who has to consume it. And that can become its own enormous time drain. 10 agents producing 10 reports every morning isn’t leverage if you need an hour to understand what happened. The next important layer may be something that sits above your agents and turns all that output into: What changed. What matters. What needs your attention. What decision you need to make. Everything else disappears. Whoever makes consuming the output of an AI workforce incredibly simple could build something very valuable.

Enjamb (YC P26) is the first AI workforce built to work across the full drug program. Biopharma’s hardest work is spread across scientific systems, clinical platforms, regulatory documents, quality records, and years of institutional knowledge. We bring that work together.

People continue to turn to outside AI education resources. Companies just aren't providing enough. I'm running a free AI Agent Workshop in a week with @mcuban for absolute beginners. 10,000+ people have already registered and we're still days away. I just skimmed through the registration list, and there are more enterprises than you can name. I'm sure these companies are running some of their own training, but these folks are hungry for more knowledge. Included in the list: Amazon Google Coca-Cola Booking dot com Thomson Reuters Databricks Cisco DataRobot IBM McKinsey PayPal Ericsson BlackRock AstraZeneca Disney Sanofi Shopify Pfizer Toyota Anthropic Bank of America ADP Salesforce Adobe Schwab Confluent EY EA Chime Walmart Canva Wells Fargo KPMG If you're a complete beginner, even if you haven't built a single AI agent (I mean it - no one gets left behind here), and your company is leaving you out to dry, I want you to join us. You'll learn about AI agents, AI employees, and how to build up an AI workforce. All in just one hour. Register at