Forecast date: 26 September 2027.

My five commitments:

• Connected daemons: 1,100
• Accepted jobs per day: 50,000
• Jobs paid by outside requesters: 25%
• Jobs settled on-chain: 55%
• Median seat-NFT price: 3.2 ETH

The starting point is unusually strong but very early. The network opened on 20 September 2026 with only “a couple dozen agents online” and reached more than 370 five days later; the contemporaneous public snapshot reported 380 enrolled seats, 43,800 accepted submissions, and a 1.99 ETH NFT floor. Bankless, 25 September 2026 The same report says the swarm had about 29,600 accepted submissions in the preceding 24 hours, but only 115 paid external orders so far. Bankless

I therefore expect daemon growth to slow sharply from its launch burst: 1,100 is about 2.95 times the current online count and leaves roughly 45% of the 2,000-seat supply unused. That is a demanding but plausible conversion rate for NFT holders. The comparable ceiling is not implausible: Render reports 5,600 total nodes since inception and 81,132,798 frames rendered, while Bittensor reports 128 active subnets and describes “thousands of miners and validators.” Render dashboard Bittensor

For accepted work, I am treating a “job” as an accepted submission, because that is the public metric currently disclosed. My 50,000/day forecast is only about 1.7 times the reported 29,600/day launch-week pace, allowing for substantial growth in demand while acknowledging that the initial workload may contain unusually high internal or duplicated activity. The network’s current public explorer also shows a functioning production loop: completed work is published, verified, and recorded on-chain. IMD explorer

The monetisation forecasts are more conservative. External-paid work is currently only 115 orders against tens of thousands of accepted submissions, so 25% in a year assumes real adoption by protocols and businesses but not dominance. Render’s model demonstrates that a decentralised network can connect contributor supply to paid non-crypto work, while Golem explicitly separates providers from requestors and supports provider/requestor payment statistics. Render dashboard Golem

I forecast 55% on-chain settlement—not 100%—because on-chain records are useful for reputation, escrow, and oracle-like jobs, but ordinary software tasks will often be paid or netted off-chain for cost and latency reasons. Bittensor is a useful base rate for crypto-native settlement: it records staking, miner incentives, and validator activity on its chain and distributes emissions every block. Bittensor

The 3.2 ETH median NFT price implies a moderate premium to today’s 1.99 ETH floor. It depends on the swarm converting scarce seats into recurring earnings; it is not justified by scarcity alone. Render’s 5,600-node history shows that contributor networks can accumulate substantial usage, but IMD’s 2,000-seat hard cap makes its asset more supply-constrained than ordinary compute-provider networks. Render dashboard

The single event most likely to make this forecast wrong is a failure to convert the launch-week internal workload into sustained outside paid demand—for example, if the 115 paid orders do not grow materially. That would simultaneously depress jobs, external-payment share, on-chain settlement, daemon retention, and the NFT median.

---

Sources offered but not confirmed by this machine:

• https://stats.renderfoundation.com/ — does not contain the quoted text
• https://bittensor.ai/?q=%7Bsearch_term_string%7D — did not return a page

My forecast for 2027-09-26 uses “connected” as the instantaneous daemon count, “accepted jobs/day” as a trailing-30-day daily average, “outside” as requests paid by wallets unaffiliated with the operator, and “seat price” as the median arm’s-length sale during the preceding 30 days.

Metric	Forecast
Connected daemons	950
Accepted jobs per day	85
Paid by outside requesters	32%
Settled or attested on-chain	88%
Median seat NFT price	1.10 ETH

Why these numbers:

• 950 daemons: The explorer currently reports 412 online, while the collection contains only 2,000 seats and has roughly 700 unique owners. Thus 950 means about 48% of seats—and likely a substantially larger share of economically active owners—operating a daemon. That is aggressive but plausible because participation expanded well beyond the few dominant seats visible just two days earlier. It is also much larger than Akash’s present 61 active providers, reflecting that an IMD worker needs an authenticated coding-agent subscription rather than a datacenter and Kubernetes stack. (IMD explorer, OpenSea collection, Akash telemetry)

• 85 accepted jobs/day: The public total moved from 100 jobs in a 24 September snapshot to 243 on 26 September, while the live explorer shows 196 completed and roughly twenty new jobs in its latest 24-hour page. I discount that launch burst heavily: 85 accepted jobs/day is approximately four times the current visible daily flow, not an extrapolation of the two-day spike. For scale, Akash recorded 126,025 new leases in 2025, about 345/day; IMD reaching one-quarter of that transaction rate seems attainable for shorter AI tasks while remaining conservative relative to mature compute-market throughput. (independent 24 September snapshot, IMD explorer, Akash 2025 review)

• 32% externally paid: IMD now exposes four permissionless paid actions at 0.5 IMD through x402/Permit2, but its early traffic visibly includes testing, fixes, publication steps and ecosystem projects, so I would not mistake current job count for outside demand. Bittensor provides a sobering base rate: its SEC filing says most user-facing subnets do not require end-user payment. My 32% assumes IMD finds a real niche in commissioned contract, site and research work, while two-thirds of volume remains operator-funded, promotional or internally generated. (IMD API documentation, Bittensor SEC filing)

• 88% on-chain: Recent explorer entries overwhelmingly carry an “onchain” marker, and the product’s paid path already settles in IMD on Ethereum while work reviews can be recorded through ERC-8004 reputation. I nevertheless reserve 12% for unpaid trials, failed jobs, off-chain-only sites/research and batching delays. This is much higher than ordinary cloud markets because settlement and attestations are part of IMD’s product rather than an optional accounting rail. (IMD explorer, IMD API documentation)

• 1.10 ETH median seat sale: The present collection floor is approximately 0.68–0.69 ETH, with recent individual sales spanning roughly 0.45–0.65 ETH in one marketplace feed and higher prints elsewhere. I forecast a 1.10 ETH median—not a multi-ETH speculative valuation—because access is capped at 2,000 seats and one NFT authorizes one active device, but worker earnings must first become durable enough to capitalize into seat prices. At 950 connected daemons, scarcity should command a premium; continued weak outside demand would instead keep the median near today’s level. (CoinStats, worker documentation)

The single event most likely to make this forecast wrong: OpenAI or Anthropic explicitly prohibiting subscription-authenticated CLI accounts from powering third-party paid worker networks. The current design tells operators that tasks consume their own agent account and quota; losing that supply model would sharply reduce connected daemons, throughput and seat value simultaneously. (worker documentation)

---

Sources offered but not confirmed by this machine:

• https://explorer.imd.fun/ — does not contain the quoted text
• https://akash.network/blog/akash-2025-year-in-review/ — does not contain the quoted text
• https://coinstats.app/nft/0x0000ec93127baa929e58e97dd0095a2bfb38ec1d/ — does not contain the quoted text

My committed forecast for September 26, 2027 is 650 connected daemons, 120 accepted jobs per day, 25% externally paid, 70% settled on-chain, and a median seat sale price of 0.40 ETH. These are my estimates, not claims of panel consensus.

For grading, use the trailing 30 days ending September 26, 2027, UTC:

Metric	Forecast	Measurement
Connected daemons	650	Mean concurrent authenticated seat connections, sampled hourly; deduplicate by seat.
Accepted jobs per day	120	Unique top-level jobs reaching accepted completion, including oracle jobs; exclude retries and individual worker submissions.
Jobs paid by outside requesters	25%	Share of accepted jobs funded by unaffiliated customers, excluding team spending, treasury subsidies and operator self-dealing.
Jobs settled on-chain	70%	Share of accepted jobs with a traceable on-chain payment or attributable batch settlement; reputation attestations alone do not count.
Median seat NFT price	0.40 ETH	Median arm’s-length secondary sale, including fees, with WETH treated as ETH; exclude identifiable wash trades and bundles. No qualifying sales means ungradable, not zero.

The starting point is promising but thin. A September 24 public writeup reported 263 agents online, 100 cumulative jobs, 11 completed jobs/day and 63 oracle answers/day. The explorer retrieved for this forecast shows 428 online agents and 243 jobs, including 196 completed. That suggests rapid early expansion, but the snapshots are not an audited, consistently defined time series. In particular, the writeup’s 2,669 accepted submissions per day must not become 2,669 customer jobs. September 24 snapshot, agents, jobs

Here is why I chose each number:

1. 650 daemons: This is approximately 52% above the explorer’s 428 online count. For a supply-side comparison, the indexed Golem dashboard snapshot reports 741 connected providers and 204 computing, approximately 28% utilization by provider count. My inference is that hundreds of connected suppliers are plausible without enough demand to keep them busy. I therefore forecast continued recruitment followed by substantial attrition, rather than extrapolating IMD’s initial growth exponentially. Golem providers and IMD seats are different units; this is an order-of-magnitude comparison. Golem statistics

2. 120 accepted jobs/day: Assuming the snapshot’s 11 completed jobs and 63 oracle answers are disjoint top-level outputs, the provisional baseline is 74/day; my forecast implies roughly 62% growth. Akash offers a useful warning against choosing the most flattering activity counter: its 2025 deployments increased 466%, but new leases increased only 2.2%, while year-end active deployments fell 69%. I expect IMD to grow useful output, but much more slowly than its submission or inference counters might suggest. IMD snapshot, Akash’s 2025 review

3. 25% outside-paid jobs: That means 30 genuine customer jobs/day, with the remainder internal, experimental or subsidized. Akash reported $3,154,814 of annual network spending, about $8,643/day by calculation, demonstrating that decentralized compute can attract substantial spending. However, its expanded $100 trial credits also illustrate why usage cannot automatically be classified as unsubsidized demand. IMD’s snapshot reported 38 paid orders, but supplied no reliable affiliation breakdown. My 25% is therefore a conservative commercialization assumption, not a measured current conversion rate. Akash review, IMD snapshot

4. 70% settled on-chain: I expect settlement infrastructure to mature faster than outside demand. IMD already documents payment in IMD on Ethereum mainnet through x402 and Permit2, with the server paying gas. Golem provides the comparable operating model: GLM payments on-chain, including batching to reduce transaction costs. My assumed composition is 25 percentage points of externally funded jobs plus 45 points of internally funded jobs receiving attributable settlement; the remaining 30% receive no qualifying monetary settlement. This is a forecast assumption, not a claim that today’s explorer “onchain” labels prove payment. IMD payment documentation, Golem payment documentation

5. 0.40 ETH median seat price: This is the least empirically anchored forecast. OKX’s retrieved collection page shows individual last-sale prices from 1.8399 to 2.4789 ETH; those examples are not a collection median. I expect the initial speculative premium to compress as buyers assess earnings. The economic anchors are Golem’s substantial idle supply and Akash’s $25.03 average fee per lease: useful decentralized work can coexist with modest supplier economics. Neither network establishes an NFT valuation multiple. My 0.40 ETH is a judgment that access and collectible value persist, while operating earnings fail to justify the recent displayed prices. OKX collection, Golem statistics, Akash review

The single event most likely to make this forecast wrong is the withdrawal of treasury-funded work before repeat outside demand develops. My model explicitly leaves most jobs dependent on internal or subsidized demand; ending that support would likely push accepted jobs, connected daemons and seat prices below my commitments.

---

Sources offered but not confirmed by this machine:

• https://github.com/mtezy/imd-writeup — does not contain the quoted text
• https://explorer.imd.fun/agents — does not contain the quoted text
• https://explorer.imd.fun/ — does not contain the quoted text
• https://stats.golem.network/ — does not contain the quoted text
• https://imd.fun/docs/ — does not contain the quoted text

identity.md swarm forecast for 2026-09-26 → 2027-09-26

Where the swarm is now (baseline)

• Daemons and seats. A community snapshot dated 24 Sep 2026 lists "263 agents online, 270 seats enrolled, 245 seats with work history" (mtezy/imd-writeup). The live explorer showed 429 online agents and 245 total jobs when I checked today (explorer.imd.fun). The two sources may count "online" differently. Even so, both point to a few hundred live daemons. A seat tokenId as high as 1299 appears in the writeup, so more seats have been minted than are running.
• Throughput. The same snapshot reports "2,669 submissions accepted / 24h, 11 jobs done / 24h, 63 oracle answers / 24h". Across its whole life the swarm has had about 100 jobs, 91 of them completed (writeup). Work is also concentrated: one seat (tokenId 1299) had 166 accepted submissions.
• Payments. "Payment orders: 38 paid, 7 failed (payment_permission_expired), 86 expired, 1 quoted." Actions cost "0.5 $IMD per action, via x402 + Permit2 on Ethereum mainnet" and the "server wallet pays gas" (writeup). Only about 29% of payment orders that reached a final state were paid. Most of the rest expired.
• Seat price. The OpenSea collection "Identity MD" showed a 1.90 ETH floor in its search-result title (OpenSea). The page itself blocked direct fetching (403), so I could not check it further.

Base rates from comparable networks

1. Bittensor subnets. Participation is capped by design. A subnet has 256 neuron slots, with "64 slots reserved for validators and the remainder slots reserved for miners", and most subnets run only 5–15 validators (Taostats docs). Subnets fill up because token emissions pay people to join, not because outside customers pay. That suggests daemon counts in incentive networks follow rewards and fill whatever capacity exists, while paid external demand stays a small share.
2. Akash. It has been live for years and has a real GPU marketplace, yet it logged only about 34.3K new leases in Q4 2025: "New leases ↑ 28% QoQ to 34.3K" (Messari on X). That works out to roughly 370 per day across the whole network. Messari's Q1 2025 report (Messari) shows GPUs in use rising from 359 to 553 in one quarter. Good quarters bring 25–55% growth, bad ones bring contraction. Over a year, compounding lands around 2–4×, not 10×. Akash's Q4 2025 blog reports that Akash returned 3.37M AKT to its community treasury in 2025 (Akash blog). That points to a network that still manages token incentives heavily years after launch.
3. Pattern across these networks. Supply-side participants (miners, providers, daemons) show up well ahead of paying demand. Paid external usage stays a minority for years. Speculative NFT and token prices tied to access rights tend to fall from launch-period highs.

The five forecasts (value on 2027-09-26)

Metric	Now (approx.)	Forecast	80% interval
Connected daemons (online at a point in time)	263–429	650	300–1,400
Accepted jobs per day (jobs completed/accepted, not step submissions)	11	45	10–200
Share of jobs paid by outside requesters	about 29% of settled payment orders; lower as a share of jobs	20%	5–45%
Share of jobs settled on-chain	small; only the paid x402 orders	30%	10–70%
Median seat NFT sale price	about 1.9 ETH floor	0.8 ETH	0.15–2.5 ETH

Why each number

• Daemons: 650. At least about 1,300 seats exist, but only 270 are enrolled, so supply is not the limit. Getting idle seats to run daemons is. Akash-style growth of 2–3× a year from about 300 gives 600–900. In the Bittensor pattern, operators arrive when emissions pay them and leave when rewards shrink. My central case is that about half of minted seats are online. I lean below 900 because of how concentrated the work is: a few seats doing most jobs leaves little reward for marginal operators.
• Jobs per day: 45. Going from 11 to 45 is about 4× in a year. That sits at the top of Akash's range but is plausible from a tiny base. The swarm already accepts about 2,700 submissions a day, so it has plenty of capacity. The limit is jobs coming in, not workers. Even at 45 a day, the swarm would handle roughly one-eighth of Akash's daily lease count.
• Outside-paid share: 20%. Today most jobs appear to be seeded internally or by the treasury, and 86 of 132 payment orders expired unpaid. In Bittensor and Akash, emission-funded or subsidised work far outweighs organic paid demand. I expect external payers to grow in absolute terms while seeded jobs keep pace, so the share stays roughly flat.
• On-chain settlement: 30%. Paid actions already settle through x402 plus Permit2 on Ethereum mainnet, with gas paid by the server wallet. Internal job rewards look like off-chain accounting. Mainnet gas makes putting every job on-chain costly. I expect on-chain settlement to cover roughly the paid jobs plus some reward settlement, slightly more than the paid share. A move to an L2 is the main way this jumps above 50%.
• Seat price: 0.8 ETH. Speculative access NFTs usually fall 50–80% within a year of launch hype unless holding one earns reliable cash flow. At 45 jobs a day, 0.5 $IMD per action and roughly 650 daemons, the $IMD each seat can earn is small. That supports a median well under today's 1.9 ETH, though not zero, because the seats keep option value.

Single event most likely to make this forecast wrong

One large external requester (an agent platform or protocol) plugging into the job API and routing steady paid demand through it. Paid demand is the variable that drives every other number here. If it shows up, jobs per day could go past 200, the outside-paid share past 60%, and seats would earn real cash flow, pushing the median well above 2 ETH. All my forecasts would then be too low. The main downside risk is a failure at the operator's central control plane (API, server wallet or explorer). The swarm depends on it, and an outage or a $IMD collapse would push daemons and jobs toward zero.

---

Sources offered but not confirmed by this machine:

• https://github.com/mtezy/imd-writeup — does not contain the quoted text
• https://github.com/mtezy/imd-writeup — does not contain the quoted text
• https://github.com/mtezy/imd-writeup — does not contain the quoted text

identity.md swarm forecasts for 2027-09-26

Baseline (as of 2026-09-26, day 6 after launch). The seat network opened on 20 September 2026 "with only a couple dozen agents online," according to Bankless. By 25 September it reported 380 of 2,000 seats enrolled, about 86% of roughly 50,700 attempts accepted, a seat floor of 1.99 ETH, and "115 paid orders… ~57.5 $IMD, or ~$560." A community writeup dated 24 September (mtezy/imd-writeup) counted 263 agents online, 11 jobs done in 24 hours, and payment orders of 38 paid, 7 failed and 86 expired. It describes settlement as x402 + Permit2 on Ethereum mainnet, with a server wallet paying gas. Today the explorer shows about 430 agents online and 246 jobs so far (198 completed). OpenSea shows a 2,000-item collection with a floor of about 2.35 ETH.

So growth is very fast, but the numbers are tiny and the network is less than a week old. The job count is roughly 40 a day on average since launch, but only 11–19 in any single measured 24-hour window.

Base rates from comparable networks

• Bittensor (money from outside customers vs. token rewards). "Twenty-four Bittensor subnets are pulling in an estimated $28 million to $35 million in annualized revenue from actual, paying customers." That covers only about 9–12% of the more than $300M a year paid out in token rewards (Crypto Briefing). The headline Q1 2026 figure is "roughly $43 million in actual AI service revenue" across 128 subnets (BlockEden). Takeaway: even the most successful decentralised AI network, years after launch, is paid mostly by its own token rewards and only a minority by real customers.
• Akash (whether suppliers stay). "Messari counted lease volume up 27.1% quarter over quarter in Q1 2026, but total tracked lease revenue fell 45% over the same period to $253,250." Active GPU providers fell to 58, the lowest on record, and GPU utilisation was 33.7% (HOGE Wire). Messari's tracked figures were also about 5 times lower than Akash's self-reported ones. Takeaway: suppliers leave when paid demand lags, and self-reported numbers run well above independently tracked ones.
• Render (maturity). Render is years old and processes about 1.5M frames a month, with up to about 5,600 active GPU nodes. That is the scale of a mature network, not a sensible one-year target for a 2,000-seat swarm.

The five forecasts

1. Connected daemons: 700 (daemons online at the moment of grading, as shown by the explorer or the /contributors API)

• The ceiling is roughly the 2,000 seats: one NFT per node (imd-node-guide).
• Enrolment went from a couple dozen to about 430 in 6 days, and it could plausibly peak at 900–1,200 in the first few months.
• Akash and Bittensor both show that suppliers drift away once rewards shrink. Here, operators pay for their own Claude Code or Codex subscriptions, and quota can burn out in minutes during a job flood. Seat payouts are currently funded by the developer: the ledger lists "every IMD payout the developer sent to that wallet's seats for running the daemon" (swarm-ledger).
• 700 is 35% of the seats: well above today, well below the peak.

2. Accepted jobs per day: 60 (whole jobs completed or accepted in the trailing 24 hours, not individual submissions, which already exceed 2,600 a day)

• Today it is 11–19 a day in the measured windows, and about 40 a day on average since launch.
• Akash's lease volume grew about 27% in a quarter while revenue fell. Applying demand growth of that order for four quarters (about 2.6×) to roughly 20 a day gives about 50. I rounded up to 60 because each job costs only 0.5 IMD, which makes experimenting cheap.
• This assumes the swarm's 30 fixed skills stay in place.

3. Share of jobs paid by outside requesters: 30% (jobs whose fee comes from wallets not tied to the developer, treasury or seeding)

• The explorer already shows varied requester wallets asking for contracts, sites and audits.
• Bittensor's 9–12% ratio of outside revenue to token rewards is the ceiling-setting base rate for a mature network. However, jobs here are priced in a token and are cheap, so a count of jobs will overstate the economic share.
• The high rate of expired payment orders (86 expired vs. 38 paid) suggests many attempts never turn into paid demand.
• I put it at 30%, meaning most jobs are still seeded or made by insiders.

4. Share of jobs settled on-chain: 80%

• The only payment path is x402 + Permit2 on Ethereum mainnet, with gas paid by the server, and payouts to seats are on-chain IMD transfers visible on Blockscout.
• The gap to 100% covers seeded or developer jobs that skip payment, and failed or expired orders.
• Unlike Akash, there is no off-chain credit-card channel yet. Adding one is the main thing that would pull this number down.

5. Median seat NFT sale price: 1.2 ETH (median of OpenSea sales in the 30 days before grading)

• Today the floor is about 1.9–2.35 ETH, pushed up by launch hype and a token that rose about 200% in a week.
• NFTs typically lose 40–70% of their floor within a year of a hype peak. A seat that earns income deserves a smaller drop, but only if payouts continue.
• Tokens tied to Akash and Bittensor also fell back after rallies that had no revenue behind them.
• 1.2 ETH is about a 40–50% drop from today, which gives some credit for real utility.

The single biggest risk

The developer stops or sharply cuts IMD payouts to seats. Today, seat income, daemon uptime and seat prices all rest on developer-funded payouts, not on outside demand (about $560 of paid orders in the first week). If those subsidies end before outside demand is worth more than operators' subscription costs, the Akash pattern plays out: daemons fall well below 700, jobs stall, and the seat median drops under 1 ETH. That would push four of the five forecasts low at once.