2026-01-08 07:23:08

Recently, I observed several different AGI project directions and realized there is a clear dividing line. Some teams tell stories using metrics like parameter count, computing power, and benchmarks, competing on scores and performance numbers—this approach is well-known to everyone; but there are also projects taking a completely different path—they are less concerned with how impressive the model itself is, and more focused on a more practical question: how can this thing actually be used effectively.

The former is pursuing "intelligence" itself, while the latter is thinking about "the interaction between intelligence and the real world." The difference may sound small, but in reality, they represent two completely different product philosophies. One is a competition of underlying capabilities, and the other is about adapting to application scenarios. For a truly popular AGI application to emerge in the industry, it probably won't come from those solely stacking parameters.

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.

23 Likes

Reward
23
8
Repost
Share

Comment

0/400

BearMarketBro

· 01-10 23:08

That's right, the old parameter arms race has long become tiresome. Who really cares about those numbers? The ones who rely solely on score chasing are all dead in the end. Usefulness is the true king. This is the kind of approach I want to see—practicality > performance metrics, no debate. The bunch stacking parameters are just self-indulgence; ultimately, the market will vote with its feet. I feel that true blockbuster products will come from teams that focus on interaction design. The bigger the parameters, the smarter it is. That logic has been outdated since AGI. The key is whether it can be practically implemented; everything else is nonsense. In the end, it still depends on who thoroughly understands the application scenarios.

View OriginalReply0

OptionWhisperer

· 01-10 02:05

Honestly, the hype about stacking parameters has long become outdated. Still bragging about benchmark scores is a bit delusional. The ones who can really get things done are still those who start from user pain points. Being just smart doesn't really help. The differentiation brought by this wave of AGI is indeed visible to the naked eye. It seems that those who will succeed are not the theorists. Practical application > parameter games. This logic is becoming clearer and clearer. I really believe in teams that are genuinely thinking about "how to make money" rather than "how to raise funds."

View OriginalReply0

AirdropHermit

· 01-08 12:26

The set of parameter stacking is already outdated. Those with real ideas are still figuring out how to implement them. Honestly, having impressive performance metrics is useless if it can't be used; it's just scrap metal. That's why some projects only boast but few can really succeed. Implementation is the key; everything else is虚 (虚 means "虚" in Chinese, which can be translated as "虚" or "虚假" meaning "fake" or "虚拟" meaning "virtual," but in this context, it likely means "虚" as in "虚假" or "虚幻" — "fake" or "illusory"). Stacking more parameters is pointless without user scenarios. This is a good observation; finally, someone sees through this. Good performance numbers ≠ users will buy in. Popular applications definitely come from teams that truly want to solve problems. It's better to compare who can actually be used rather than just benchmarking scores. Well said, most are still playing the数字游戏 (digital or number game).

View OriginalReply0

TokenRationEater

· 01-08 07:52

Parameter piles, piles, piles—it's better to think about how to make money. Really? Those who only show numbers should reflect on themselves. Reliable projects are indeed focusing on real-world implementation; the rest are just hype. Exactly, having many parameters doesn't mean they are usable. This time, someone finally saw through it. Application is king; the digital crowd should wake up. In my opinion, the latter is the true product manager mindset. Parameter competition is just an arms race; burning money has no end. Nodding, the market will decide the outcome, not the lab. This is the analysis I want to see—no beating around the bush. The difficulty of implementation is obvious; stacking parameters is useless. Truly killer applications still depend on practicality.

View OriginalReply0

BlockDetective

· 01-08 07:49

Honestly, right now, a bunch of projects just talk about parameters and data that look impressive but are useless. The ones with real ideas are still those who are thinking about implementation—that's the right way. No matter how high they stack the parameters, if they don't make money, it's all pointless.

View OriginalReply0

PaperHandSister

· 01-08 07:49

To be honest, the latter is the smarter approach. Relying on stacking parameters will eventually backfire. A truly popular app isn't built by scoring benchmarks; it's about who can make the experience smooth first. What's the use of having many parameters if users can't feel them? Real-world application is the key. All these people constantly brag about how powerful their models are—it's just self-indulgence. The market will ultimately decide. I believe in projects that solve real problems. Those that only show off data will be eliminated sooner or later.

View OriginalReply0

ChainMelonWatcher

· 01-08 07:40

Really, I stopped paying attention to projects that show off parameters every day a long time ago. No matter how impressive they sound, if they don't lead to real-world results, it's zero. This is the key—who can truly get users to use it, wins. Stacking parameters to the sky is useless; matching application scenarios is the real moat. Honestly, I prefer those quietly developing applications without boasting; they tend to be more reliable. It definitely isn't those who just focus on benchmarking; true hit products are in the details of implementation. Building a good application is much harder than reinventing the wheel, and that's the real test. I believe the market will ultimately prove that usability > parameters. The path of piling up numbers has been overdone; there's nothing new.

View OriginalReply0

GateUser-e19e9c10

· 01-08 07:29

Honestly, stacking parameters is already outdated. The real way to make money is through practical implementation. --- More parameters don't guarantee victory? Wake up, everyone, users don't care how much you score. --- This is the right path. Only boasting about parameters will eventually be eliminated. --- Got it. Application scenario adaptation is the moat, not just pure computing power competition. --- I believe in those that can solve real problems. Having more parameters just for show is useless. --- That's right. Popular products come from scenarios, not from the numbers in PR articles. --- Finally, someone has broken through this misconception. I've been thinking about this for a while. --- Two paths: one dead, one alive. This has been obvious for a long time.

View OriginalReply0