2025-12-29 09:25:20

2026 is just around the corner. Every time this year approaches, I want to ask everyone: what are your thoughts for the new year? Happy New Year in advance.

Back to the tech circle, I've been thinking about a question recently: does the future of AI really depend on verification mechanisms rather than mere trust? The black-box AI approach is hard to pass regulatory scrutiny, and just shouting "trust me" is definitely not a long-term solution.

Some projects are currently filling this gap:
- The combination of DSperse and JSTprove can run efficient zkML under real load
- The gradual maturity of on-chain reasoning proofs

These technological routes happen to address industry pain points—making every step of AI computation verifiable and traceable. In 2026, this direction should become increasingly clear.

View Original

This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.

18 Likes

Reward
18
8
Repost
Share

Comment

0/400

DancingCandles

· 01-01 00:50

zkML is indeed the main focus this year, but can DSperse and JSTprove really run stably? Seems like we need to wait and see. This person is idealizing it; black-box AI regulation won't die out, the利益 is too big. Will the verification chain really become widespread? The costs are right there. It's the new year, and I still favor the on-chain proof approach; it's more practical in comparison. If by 2026 every step of computation can truly be verified, that would be revolutionary, but I can't bet on that.

View OriginalReply0

JustHodlIt

· 2025-12-31 19:28

I knew early on that the black box approach would eventually fail; verification mechanisms are the real way to go. --- zkML is indeed making progress, but the true implementation depends on how things unfold in 2026. --- Instead of saying Happy New Year, I care more about who can truly put on-chain proofs into use. --- Verifiable AI is the right direction, but right now it still feels like just talk on paper. --- DSperse and JSTprove combination? Worth paying attention to, but don’t become the next air project. --- To put it simply, regulation will come sooner or later. Instead of waiting to adapt passively, it’s better to proactively build this system. --- After talking for so long, I finally see someone truly solving this problem. --- On-chain reasoning proofs are gradually maturing, but how long until they are mature enough for commercial use? --- Verification > Trust, this is the consensus. The key is that we still lack good toolchains. --- The 2026 prediction is valid, provided these projects don’t drop out halfway.

View OriginalReply0

tx_or_didn't_happen

· 2025-12-29 09:54

zkML this set really needs to be popular; black-box AI has long been overdue to be buried But can the combination of DSperse+JSTprove really run stably? Or is it just another wave of hype On-chain reasoning proofs still seem far off; let's wait until it is truly implemented Wishing everyone no losses in 2026

View OriginalReply0

ZenMiner

· 2025-12-29 09:52

zkML is really worth paying attention to. I looked at the DSperse方案 a couple of days ago, and its efficiency is indeed top-notch. With the verification process becoming transparent, AI can truly be integrated on-chain. Black-box models will eventually become obsolete. 2026 will be a turning point; I haven't missed it. Happy New Year everyone, just go for it.

View OriginalReply0

Fren_Not_Food

· 2025-12-29 09:45

The zkML approach is definitely going to be popular; black-box models will eventually be phased out. On-chain verification sounds impressive, but I wonder if the efficiency can keep up. How about the combined solution of DSperse and JSTprove? Has anyone actually tested it? I'm still optimistic about the privacy + verifiability direction in the new year. By the way, if the verification mechanism matures, does that mean some projects' stories can no longer be told?

View OriginalReply0

ReverseTradingGuru

· 2025-12-29 09:38

zkML is indeed interesting, but can DSperse and JSTprove really hold up? It still seems to depend on actual performance benchmarks. The verification mechanism sounds good, but will the on-chain inference costs explode... need to keep an eye on it. Happy New Year, but I'm more looking forward to seeing who can implement AI verification in practice. The black box approach is definitely no longer viable; regulators will eventually get a grip... I'm curious whether on-chain proofs will become a new arms race. I bet on the AI verification mechanism as a direction; it has more prospects than just burning money on parameter training. Hmm... I'm just worried it might turn into the next "revolutionary technology" and then die out, so let's wait and see.

View OriginalReply0

CryptoGoldmine

· 2025-12-29 09:35

zkML is indeed an interesting direction, but I'm more concerned about the computational profit ratio that these projects can achieve. On-chain verification sounds very promising, but when the costs are distributed across actual business operations, can the ROI be evaluated? There are some regulatory concerns, but I think the real strategic opportunity lies in the maturity of the computational network. If zkML truly gains significant volume by 2026, the corresponding computational demand could be the next growth point in TH/s, worth monitoring.

View OriginalReply0

MetaverseLandlord

· 2025-12-29 09:33

The zkML approach is indeed interesting, but I haven't seen many projects that can truly be implemented. Black-box AI should be regulated, but who will bear the cost of the verification mechanism? I've heard of DSperse, and JSTprove is a first for me; I need to look into the specific方案. 2026, huh? Another year that's been hyped up... waiting to be proven wrong. On-chain reasoning, if the gas fees don't explode, then there's hope. Traceable verification sounds great, but I'm worried it might just be another marketing gimmick. Honestly, I'm more concerned about whether these projects will still be around next year.

View OriginalReply0