
Kaspa DAGKNIGHT could replace GHOSTDAG with responsive consensus. Learn how the upgrade works, its status, benefits and mainnet risks.
Author: Kritika Gupta
14th September 2026- Kaspa’s community account restated an old research result this week, and the crypto crowd noticed. The account @kaspaunchained called DAGKNIGHT a proposed replacement for GHOSTDAG. It also added three plain words: still in development.
High Signal Summary For A Quick Glance
Unique Human | Changer ($CNG)
@changerofficial
@kaspaunchained Love seeing Kaspa continue pushing consensus innovation.
DAGKNIGHT is a proposed replacement for GHOSTDAG, Kaspa’s current consensus protocol. Blocks take time to reach other miners. GHOSTDAG accounts for that with a preset delay assumption, which influences confirmation times even when the network is running smoothly. DAGKNIGHT https://t.co/SpbIlfZTk9
01:34 PM·Sep 14, 2026
Ben Ralos⚡
@AmpVoltage
@kaspaunchained Kaspa is technically exceptional but idk
DAGKNIGHT is a proposed replacement for GHOSTDAG, Kaspa’s current consensus protocol. Blocks take time to reach other miners. GHOSTDAG accounts for that with a preset delay assumption, which influences confirmation times even when the network is running smoothly. DAGKNIGHT https://t.co/SpbIlfZTk9
01:08 PM·Sep 14, 2026
High attention and emotional sentiment detected.
So this was not a launch. Instead, it was an explainer about Kaspa DAGKNIGHT, the consensus upgrade the project has floated since 2022. The post drew light traffic, roughly 3,100 views and 35 likes at fetch time.
DAGKNIGHT is a proof-of-work protocol that orders blocks without a hardcoded latency assumption. Its authors call it the first permissionless protocol with no built-in bound over network latency. In short, it adapts to real conditions instead of a fixed guess.
Kaspa runs on a blockDAG rather than a single chain. Bitcoin allows one parent per block, so parallel blocks get orphaned. A blockDAG keeps those parallel blocks and sorts them into one order later. That design lets Kaspa push 10 blocks per second without wasting work.
Two names sit behind the design. Yonatan Sompolinsky founded Kaspa and now works at Harvard. Michael Sutton leads core development. Together they published the DAG KNIGHT paper in 2022. The protocol also tolerates an attacker holding up to 50% of network hashrate.
Kaspa mainnet runs GHOSTDAG today. That protocol uses a fixed parameter called k, and k comes from a worst-case guess about how long blocks take to spread across miners.
Because that guess sits inside the code, the network waits as if it were slow even when it runs fast. After the Crescendo upgrade in May 2025, mainnet set k at 124 and moved to 10 blocks per second.
Kaspa DAGKNIGHT flips that logic. It reads the DAG and infers the smallest k that still covers a majority of blocks. As a result, confirmations tighten when the network is healthy, and the chain needs no new hard fork to adjust.
The idea did not appear overnight. It grew from SPECTRE in 2016 and the PHANTOM and GHOSTDAG paper in 2018. Each step pushed toward faster ordering on a blockDAG.
The work lives on the public dagknight branch of rusty-kaspa. In early September, developer coderofstuff wired per-network activation and added a dedicated testnet-13 genesis. There, DAGKNIGHT runs on while mainnet stays off.
That change merged through pull request 1120 on 6 September 2026. Still, it does not launch a public testnet, and it does not touch mainnet. The first testnet-13 build even lists no DNS seeds.
A follow-on pull request 1127 goes further. It proposes a bounded variant that already deviates from the original paper. So the shipped protocol may differ from the 2022 design.
GHOSTDAG vs DAGKNIGHT
Marketing copy sometimes claims instant finality. The paper does not. Confirmation stays probabilistic, and each wallet picks its own safety margin, much like waiting six Bitcoin blocks instead of one.
Sompolinsky has also cooled the speed hype himself. In an August 2026 interview, he said DAGKNIGHT is “not faster per se.” Instead, it confirms at whatever latency the network actually shows.
The label “parameterless” needs care too. Clients still choose a conservative delay bound, and a global cap remains for difficulty and pruning. So DAGKNIGHT trims assumptions rather than removing them.
The tweet did not move the market. KAS traded near $0.0335 on 14 September, with a market cap around $930 million, according to CoinGecko. Daily volume sat near $7 million, an ordinary figure.
The session ran soft red, down roughly 3% over 24 hours. Still, KAS held a gain of about 31% across the prior 30 days. The token also sits far below its July 2024 record near $0.207. This article is not financial advice.
Kaspa-native accounts mostly frame this as building, not shipping. One popular post put it bluntly: code, then testnet, then proof, then mainnet. That sequence sets expectations against date hype.
Not everyone agrees on the pace. Some voices still talk as if activation sits weeks away. A separate chain called Hoosat even claims it ran a public DAGKNIGHT fork first. Still, that is a competitor talking point, not Kaspa mainnet evidence.
Search noise clouds the topic too. A memecoin named DagKnight Dog crowds the term on X. So readers hunting the protocol often land on an unrelated token instead.
For now, the roadmap order looks clear even if the calendar does not. Kaspa shipped Toccata covenants on 30 June 2026, and DAGKnight follows next in the official lore.
Yet the mainnet date stays blank. KIP-2 still carries the Proposed label, no activation score exists, and an older Q3 2026 target has quietly lapsed. So watch the dagknight branch and the KIP-2 status for the real signal, not the meme layer.
Our Crypto Talk is committed to unbiased, transparent, and true reporting to the best of our knowledge. This news article aims to provide accurate information in a timely manner. However, we advise the readers to verify facts independently and consult a professional before making any decisions based on the content since our sources could be wrong too. Check our Terms and conditions for more info.