Tokenomics Update: One Quarter of On-chain data
Part of our ongoing commitment to build the Cosmos Hub roadmap in public, as laid out in From Chaos to Stability to Growth . A few weeks ago we shared the first output of our tokenomics research with Gauntlet, “ Where Does One Day of ATOM Emissions Go ?” , which traced how a single day of ATOM emissions moves through the network and found that emissions contribute very little to actual sell pressure. This update goes a level deeper and asks a harder question. Across all of ATOM’s sell pressure, who is doing the selling, and how much? This is an early look at Gauntlet’s Phase 1 sell-pressure analysis, now extended across the full first quarter. Over sixteen consecutive weekly windows (January through April 2026), the model attributes 25,523,746 likelihood-weighted ATOM of sell pressure to the holder cohorts that initiated it. These findings are still being finalized. We’re sharing the direction now and will follow with the complete, validated report. What we’re measuring We never assume a transaction is a sale. Every movement of ATOM out of a tracked cohort is sorted into a route, and each route carries a likelihood : the share of that flow we treat as genuine selling.…
Excerpt (1196 of 5545 characters). Read the whole post on the forum ↗
RoboMcGobo:It’s good that we did. Anecdotal assumptions were that inflation (via staking rewards and validator commissions) were driving the bulk of sales activity.
In fact, selling is concentrated among a small set of large actors moving in bursts, not broad-based, persistent, inflation-driven selling from everyday stakers. That distinction matters for how we think about inflation, emissions, and which levers the Hub actually has.
Over the years, many networks assumed that aggressively cutting inflation would lead to a major price increase. The evidence showed that always the opposite happened but projects kept believing in that flawed idea. It is great that finally the Cosmos hub has proven via this research how wrong all those projects were. In fact, let’s remember that even the Cosmos hub was partially a victim also, the max inflation parameter was previously halved by half and further actions were planned but fortunately not approved. Looking forward to the next steps of the research
While the sale pressure shows for the past quarter very little sale from the validators and lots of retail being scared out and also large investors just shows that most of the market thinks believes the current market sentiment. The inflation is still a little high to be competitive with other PoS networks. I believe 8% on the high end and 4% on the low end is optimal. Furthermore it makes sense that people who stake and collect inflation to secure the network are probably selling as little as possible over the past quarter.
You all are the ones making assumptions off of 3 months of data.
Reading the OP, the thing that gets glossed over in most Cosmos discussions about Tokenomics Update: One Quarter of On-chain data is the heterogeneity between zones. The IBC numbers and the staking-ratio numbers are usually aggregated across all hub-connected chains, which papers over the fact that the median zone has a totally different fee-market, validator-set, and slashing-history than the Hub. So when someone proposes a parameter change “for Cosmos”, in practice the impact is concentrated on whichever subset of zones run a custom ICS or a custom fee-token. From an applied-stats angle, the empirically interesting question is how correlated validator slashing events are across zones once shared-security and ICS-v2 are widespread. The early data from Neutron and Stride is small-sample, but the correlation already looks higher than the naive independence assumption would predict — same operator, same uptime issues, same client-version bug. If the proposal here doesn’t account for that correlation explicitly, the tail risk on the consumer chain is going to be worse than the per-validator-slashing-rate math suggests. One concrete thing I’d want before this goes to vote: a…
Excerpt (1193 of 1508 characters). Read the whole post on the forum ↗