Proof of Space 2.0: Community Q&A Summary

Chia Network · Community Call · April 16, 2026

Featuring: Bram Cohen and Dr. Nicholas Halper (Dr. Nick)

Chia’s Proof of Space 2 (PoS2) is a fundamental upgrade to the consensus mechanism of the Chia blockchain. It addresses plot compression and grinding in PoS1.

For the full technical details, see CHIP-48 (proof of space algorithm) and CHIP-49 (timelines and parameters) in the Chia Network GitHub repository, as well as the PoS 2.0 Overview Video.

Questions:

What is Proof of Space 2.0 and why is it being introduced now?
What are CHIP-48 and CHIP-49, and what can people get out of each?
How does PoS2 actually resist compression attacks?
Why is the plot size changing, and what happened to K values?
What are plot groups and meta-groups, and why do they matter?
What is plot “strength” and does plotting at higher strength earn more rewards?
What security threats is PoS2 specifically designed to stop?
How does PoS2’s compression resistance compare to what Mad Max and Dr. Plotter achieved in PoS1?
Is Chia using AI to help harden the protocol, and what about quantum resistance?
Will farmers need to replot, and how painful will it be?
When is the hard fork, and how long is the PoS1 phase-out period?
Can I plot PoS2 plots right now?
What else is coming in Chia 3.0 beyond PoS2?
What’s happening to the pooling protocol?
Will plotting be supported on AMD GPUs, not just NVIDIA CUDA?
Can tape drives be used for farming with PoS2’s advance-notice system?

Overview

Q: What is Proof of Space 2.0 and why is it being introduced now?

A: PoS2 is an iteration on the PoS1 algorithm that adds a new layer we believe obliterates plot compression attacks. The whole point of proof of space is that people shouldn’t be able to use extra computational power to claim credit for more space than they actually have. A few researchers, including Nick on this call, found techniques to do exactly that in PoS1. PoS2 closes those gaps.

On top of the security improvements, it also makes the costs of plotting and farming substantially lower from an operational standpoint. We’re releasing it now because it’s taken us this long to figure it out. What we have is very simple and beautiful, but not trivial to arrive at. This is new computer science research, and we’ve got it worked out now.

Q: What are CHIP-48 and CHIP-49, and what can people get out of each?

A: CHIP-48 is focused on the proof of space itself, how we take data, put it onto disk, retrieve it, and use it for proving and validation. It’s very technical and probably only a handful of people will dive deep into it. CHIP-49 is more about the plot filter: how often proofs get called by the network, what the disk load should be, and so on. Those parameters are much more open-ended and the community can give meaningful feedback on them.

Technical Design

Q: How does PoS2 actually resist compression attacks?

A: The core innovation is decoupling the quality string from full proof data. In PoS1, compressed plots had to reconstruct data at every single challenge, pushing the protocol toward proof-of-work territory. In PoS2, proof fragments determine proof quality directly, full reconstruction only happens when you actually win, and you have up to 20 seconds to do it. A Raspberry Pi 5 finishes in under 8 seconds.

The chaining filter is the key mechanism. For each challenge, the harvester builds a 16-link chain from two sets of proof fragments. For an honest farmer, this requires roughly 60,000 hashes. But if an attacker drops even one bit, they introduce false positives at every link, forcing approximately 536 million hashes per chain. That’s about 8,000x more work. At two bits dropped, the hash count exceeds 23 trillion. It’s completely infeasible.

By the numbers:

– Honest farmer: ~60,000 hashes per challenge

– 1-bit drop attack: ~536 million hashes per challenge

– 2-bit drop attack: ~23 trillion hashes (infeasible on any hardware)

– Overall leverage gap vs. GPU attacks: 1,000x

Q: Why is the plot size changing, and what happened to K values?

A: PoS2 only allows K28 instead of K32. The K value is the bit-width of the seed values, K32 plots had about 100 GB because of the seven-table structure. PoS2 only stores one table (T3), making plots about seven times smaller already. K28 shrinks them further, landing at roughly 1 GB per plot.

But thousands of 1 GB files is a headache for operating systems. So PoS2 introduces plot groups, you can bundle plots into a single grouped file anywhere from 1 GB to a TB or more. Effectively, K values are becoming a remnant. What matters is the group size you choose.

Q: What are plot groups and meta-groups, and why do they matter?

A: Plot groups are the most important lever for controlling disk load. All plots in the same group respond to a challenge simultaneously, requiring just two seeks and two reads total for the entire group, regardless of size. This shifts disk I/O from seek-bound to bandwidth-bound, and keeps activity very low as group size grows.

Meta-groups guarantee that plots in different groups on the same disk never respond to the same challenge simultaneously. This prevents peak load spikes where you’d occasionally need double or triple the throughput at once. It’s especially useful for very large drives. The system also provides at least 40 seconds of advance notice before a challenge, giving spun-down drives time to wake up.

Q: What is plot “strength” and does plotting at higher strength earn more rewards?

A: Strength lets you trade more plotting work upfront for less frequent disk access during farming. At base strength, challenges hit roughly every 45 minutes. At maximum strength (+4), that drops to approximately once every 21 hours. The system gives you at least 40 seconds of advance warning so you can spin up a sleeping drive, respond, and spin it back down.

No, higher strength does not earn more rewards. You earn the same amount. It just shifts the cost balance between plotting and farming. Depending on your hardware setup, tilting that balance one way or the other may be more economical for you, but your expected earnings are identical.

Security & Attacks

Q: What security threats is PoS2 specifically designed to stop?

A: There are two main threats. The first is compression, not information-theoretic compression that everyone does, but attacks where someone trades massive compute for reduced storage, gaining an unfair edge. PoS2 is algorithmically highly resistant to this.

The second is grinding, using a GPU to rapidly generate and discard plots when a challenge comes in, effectively farming without storing anything. You can’t completely prevent it, but you can make it extremely uneconomical. In PoS2, grinding attackers must redo the full reconstruction at every challenge, while honest farmers get their quality string before reconstruction is needed. That asymmetry makes grinding deeply impractical.

Both attacks historically drive centralization: someone gets a small edge and everyone flocks to their proprietary tool. PoS2 is designed to prevent either path from being viable.

Q: How does PoS2’s compression resistance compare to what Mad Max and Dr. Plotter achieved in PoS1?

A: They’re not really comparable. PoS1 compression wasn’t literal bit-dropping, it stored alternative data structures that let you rapidly reconstruct the actual proof with extra compute. Mad Max and Dr. Plotter achieved ~50% size reduction that way. In PoS2, we’re talking about literal bit-drops. A best-case GPU attack (using a 5090 at 300+ watts) could achieve roughly 0.73 bits dropped, about 2.5% compression. Once you factor in hardware cost and implementation overhead, the real-world savings drop below 1%. It’s essentially not worth pursuing.

Q: Is Chia using AI to help harden the protocol, and what about quantum resistance?

A: Our security team has been building tooling to find and get ahead of standard security vulnerabilities, that’s what the 2.7.0 release was about. AI is quite good at finding conventional security problems. The proof of space analysis, however, is deeply involved math, and AI hasn’t turned up anything interesting there.

On quantum resistance: the proof of space primitive itself already uses secure hashes, which are extremely quantum resistant. The potential vulnerability is in our BLS signing keys. Quantum-resistant signature schemes exist but produce very large signatures, a serious cost on a blockchain. We’re keeping an eye on the field and will upgrade if needed, but we’re not moving on it right now.

Transition & Farming

Q: Will farmers need to replot, and how painful will it be?

A: Yes, everyone will need to replot. For people currently running compressed plots, it will actually be a fairly quick process, plots are now small (around 1 GB), and plotting itself won’t be too bad. You also have a configurable trade-off: cheap initial plotting means more disk seeks while farming, and vice versa. Most people can just use the recommended defaults and be fine. We’ll also provide a background utility that replots automatically, you just say “go” and it handles it while you continue farming PoS1.

Q: When is the hard fork, and how long is the PoS1 phase-out period?

A: The hard fork is currently scheduled for November 2026. On activation day, PoS2 plots become valid and PoS1 plots begin a gradual phase-out, not by earning fewer rewards per win, but by having their win probability decline on a linear schedule until it eventually reaches zero.

The phase-out window will likely be 256 days, but is still open for community input via CHIP-49. The goal is to avoid a single frantic replot deadline. You’ll hit your personal crossover point, where replotting to PoS2 becomes ROI-positive, somewhere in the middle of that window depending on your current compression level.

Q: Can I plot PoS2 plots right now?

A: Not on mainnet. Testnet is planned soon. You can currently plot a single plot to verify the algorithm works, but the tooling for grouped plots, which is what you’ll actually want, is still in development. Watch the Chia Discord and social channels for testnet announcements.

Q: What else is coming in Chia 3.0 beyond PoS2?

A: Mostly security fixes and consensus cleanups. The most material change for developers is a new sha256tree opcode, which will be useful for smart coin work. Cost calculations are also being updated, which means wallets will need to upgrade to correctly declare transaction costs. Everyone will need to upgrade, but it should be a smooth process. If you haven’t already, upgrade to 2.7.0 [now](https://www.chia.net/downloads/), it addresses several vulnerabilities identified through AI-assisted security analysis.

Pools, GPU & Ecosystem

Q: What’s happening to the pooling protocol?

A: We’re releasing a new pooling protocol alongside PoS2. It’s mainly an engineering cleanup, the original protocol was one of the most sophisticated smart contracts written at the time, built under deadline pressure, and it shows. The new version uses modern singletons and is much cleaner.

A key change: payouts are moving from pay-to-key to pay-to-singleton. This lets pools do things like transfer ownership, collaborate with other pools, and defer payments. From a practical standpoint, if you operate a pool, PoS2 is effectively a new pool, your existing customers will choose whether to move to your PoS2 pool. Some pool operators have already signaled they won’t support PoS2, so we’ve designed the transition so customers can freely choose any pool when switching over.

Q: Will plotting be supported on AMD GPUs, not just NVIDIA CUDA?

A: Probably not via OpenCL, but likely via Vulkan. The plan is to support as many GPU types as possible, potentially including Intel, though onboard iGPUs may not see much benefit since the hashing already uses hardware acceleration on the CPU. PoS2 is actually simpler to implement on GPU than PoS1 was. The core operations are essentially hash-and-sort repeated, compared to the complex matching kernel that Dr. Plotter spent significant time optimizing for PoS1.

Q: Can tape drives be used for farming with PoS2’s advance-notice system?

A: Almost certainly not economically viable. The core differentiator is seek time: hard drives seek in ~10–20 milliseconds, tape drives take ~10 seconds. At each challenge, a farmer must do two seeks. That’s 20 seconds of seeking on a tape drive, within a 30-second response window, leaving almost no margin. Bursty challenge patterns would make this practically impossible.

We’re still doing some analysis on the plot filter to ensure it supports hard drive spin-down while making tape drives uneconomical, but the base proof of space algorithm itself is unchanged.