Has AI made life more convenient? Balaji and A16z discuss how to shorten the time cost of verifying AI content?

Generative AI significantly enhances the speed of information output, but it also raises an overlooked concern: "Verification Gap (Verification Gap)". From former Coinbase CTO Balaji to OpenAI co-founder Karpathy, and venture capital giant a16z, all warn: "The verification costs of AI output are becoming the biggest bottleneck and risk of the new era."

Balaji: The "verification cost" for AI users has become a real bottleneck.

Former Coinbase CTO Balaji pointed out in a post last month that the usage process of AI can be divided into two stages: "input prompt (prompting)" and "verify output (verifying)."

The former can be done by anyone, just typing a few lines; but the latter is more difficult, requiring professional knowledge, patience, and logical thinking to determine whether the AI has made an error or experienced a "hallucination (".

AI PROMPTING → AI VERIFYING

AI prompting scales, because prompting is just typing.

But AI verifying doesn’t scale, because verifying AI output involves much more than just typing.

Sometimes you can verify by eye, which is why AI is great for frontend, images, and video. But…

— Balaji )@balajis( June 4, 2025

He stated that this kind of gap is easy to handle in images or videos, as the human eye is inherently good at judging visual content. However, once faced with code, technical articles, or logical reasoning, the verification work becomes quite tricky:

The most important question when using AI is, how can I verify that the output of this AI model is correct in a cost-effective manner? We need other tools or products to verify content in areas beyond the visual domain.

He added: "For users, AI verification is just as important as AI prompts."

Karpathy: AI has accelerated creation, but has not reduced the verification process.

OpenAI co-founder and father of Tesla's autonomous driving, Andrej Karpathy, further extends Balaji's viewpoint, pointing out that the essence of creation is a process that has two stages and is repeated: "generation )generation(" and "discrimination )discrimination(": "You make a stroke )generation(, and you also have to step back and think whether this stroke truly improves the work )discrimination(."

He believes that large language models )LLM( greatly compress the time cost of "generation," allowing users to instantly obtain a large output, but do not help at all in reducing the cost and workload of "judgment." This is especially serious for code:

LLMs can easily generate dozens or even hundreds of lines of code, but engineers still need to read, understand, and check all the logic and potential errors line by line.

Karpathy stated that this is actually the most time-consuming task for most engineers, which is known as the "Verification Gap )Verification Gap(". AI has accelerated the creation process, but this time cost is directly transferred to the verification.

) From financial advisors to secretaries, the trust challenges of AI agents: Can we trust the autonomous decisions of artificial intelligence? (

a16z: The trust crisis of the generative era must be filled by cryptographic technology.

The well-known venture capital firm a16z approaches from the institutional and industrial levels, believing that AI technology will accelerate the proliferation of "fake information" because the barriers to generation are low and verification is difficult, leading to the internet being flooded with a large amount of counterfeit content. a16z advocates that trust should be engineered, and the solution is to introduce cryptographic technologies, such as:

Perform cryptographic processing on data produced by AI in stages )hashed posts(

Create using blockchain-verified identities )crypto IDs(

Establish a source-reliable content chain through the public and traceable nature of on-chain data.

These practices not only ensure that information remains immutable and verifiable but also establish a defense for the credibility of content in the AI era, with the potential to become an important intersection between cryptography and the field of AI.

)Messari Special Report: How does the Mira protocol enable AI to be more honest through a decentralized consensus mechanism?(

From prompt words to verification capabilities, the new literacy and demands of the AI era have taken shape.

Currently, generative AI brings exponential growth in information productivity, but without equally efficient verification capabilities to complement it, users may instead find themselves trapped in a dilemma of time-consuming operations and misinformation pollution.

Therefore, the core skill in the current AI era is no longer just about writing precise prompts, but rather the ability to effectively and cost-efficiently verify the outputs of AI. Whether through mutual reviews of AI models or professional verification tools, it is particularly important.

Does this article suggest that AI has made life more convenient? Balaji and a16z discuss how to shorten the time cost of verifying AI content. Originally appeared in Chain News ABMedia.