a16z founder: In the age of Agents, what truly matters has changed

Original Title: Marc Andreessen introspects on Death of the Browser, Pi + OpenClaw, and Why "This Time Is Different"

Original Translation: FuturePulse

Signal Source: This is a16z founder Marc Andreessen's latest interview on the Latent Space podcast. He is a renowned American internet entrepreneur and one of the key figures in the early development of the internet; after founding a16z, he became a representative figure among top investors in Silicon Valley. The entire conversation revolves around the history and latest trends in AI development, making it very worthwhile to read.

I. This round of AI is not a sudden emergence, but the first comprehensive "start working" after 80 years of technological marathon

• This round of AI is not a sudden emergence, but the result of an 80-year technological marathon.

• Marc Andreessen directly refers to the present as "80-year overnight success," meaning that the sudden explosion in the public eye is actually the concentrated release of decades of technological reserves.

• He traces this technological thread back to early neural network research and emphasizes that the industry has now accepted the judgment that "neural networks are the correct architecture."

• In his narrative, the key nodes are not single moments but a series of accumulations: AlexNet, Transformer, ChatGPT, reasoning models, and then agents and self-improvement.

• He particularly emphasizes that this time, it is not just text generation that has become stronger, but four types of functionalities have emerged simultaneously: LLMs, reasoning, coding, and agents/recursive self-improvement.

• He believes that "this time is different" not because the narrative is more compelling, but because these capabilities have begun to work on real tasks.

II. The agent architecture represented by Pi and OpenClaw is a deeper software architectural change than chatbots

• He describes agents very specifically: essentially "LLM + shell + file system + markdown + cron/loop." In this structure, LLM is the core of reasoning and generation, the shell provides the execution environment, the file system saves the state, markdown makes the state readable, and cron/loop provides periodic wake-up and task advancement.

• He believes the importance of this combination lies in the fact that, aside from the model itself being new, all other components are parts of the software world that are already mature, understandable, and reusable.

• The state of the agent is saved in files, allowing for cross-model and cross-runtime migration; the underlying model can be replaced, but memory and state are still retained.

• He repeatedly emphasizes introspection: agents know their own files, can read their own states, and can even rewrite their own files and functions, moving towards "extend yourself."

• In his view, the real breakthrough is not just that "the model will answer," but that agents can utilize existing Unix toolchains to harness the potential capabilities of the entire computer.

III. The era of browsers, traditional GUIs, and "human-clicked software" will gradually be replaced by agent-first interaction methods

• Marc Andreessen has clearly stated that in the future "you may no longer need a user interface."

• He further points out that the main users of software in the future may not be humans, but "other bots."

• This means that many interfaces designed for human clicking, browsing, and form-filling will degrade into the execution layer called upon by agents.

• In this world, humans are more like the ones who set goals: telling the system what they want, and then agents call services, operate software, and complete processes.

• He connects this change to a larger software future: high-quality software will become increasingly "abundant," no longer a scarce product handcrafted by a few engineers.

• He also predicts that the importance of programming languages will decline; models will write programs across languages and translate between them, and in the future, humans may care more about explaining why AI organizes code in a certain way rather than sticking to a specific language.

• He even mentions a more radical direction: conceptually, AI may not only output code but also directly output lower-level binary code or model weights.

IV. This AI investment cycle is similar to the 2000 internet bubble, but the underlying supply and demand structure is different

• He recalls that during 2000, the crash was largely not due to "the internet not working," but rather due to overbuilding of telecommunications and bandwidth infrastructure, with fiber optics and data centers being laid out in advance, followed by a long period of digestion.

• He believes that today there are indeed concerns about "overbuilding," but the current investors are mainly large companies like Microsoft, Amazon, and Google with ample cash, rather than highly leveraged fragile players.

• He specifically points out that now, as long as an investment forms a runnable GPU, it can usually be quickly converted into revenue, which is different from the large amounts of idle capacity in 2000.

• He also emphasizes that what we are using now is actually a "sandbagged" version of technology: due to insufficient supply of GPUs, memory, data centers, etc., the potential of models has not been fully released.

• In his judgment, the real constraints in the coming years will not only be GPUs but also the interlinked bottlenecks of CPUs, memory, networks, and the entire chip ecosystem.

• He juxtaposes AI scaling laws with the past Moore's Law, believing that they not only describe patterns but also continuously stimulate capital, engineering, and industrial collaboration.

• He mentions a very unusual but important phenomenon: as the speed of software optimization increases, certain older generation chips may even become more economically valuable than when they were first purchased.

V. Open source, edge inference, and local execution are not marginal, but part of the AI competitive landscape

• Marc Andreessen firmly believes that open source is very important, not just because it is free, but because it "teaches the whole world how it is done."

• He describes open-source releases like DeepSeek as a "gift to the world," because code + paper will quickly spread knowledge and raise the baseline of the entire industry.

• In his narrative, open source is not just a technical choice but may also be a geopolitical and market strategy: different countries and companies will adopt different openness strategies based on their own business constraints and influence goals.

• He also emphasizes the importance of edge inference: in the coming years, centralized inference costs may not be low enough, and many consumer-level applications cannot bear the long-term high costs of cloud inference.

• He mentions a recurring pattern: models that seem "impossible to run on a PC" today often can indeed run on local machines just a few months later.

• Besides cost, factors promoting local execution include trust, privacy, latency, and usage scenarios: wearable devices, door locks, portable devices, etc., are more suitable for low-latency, on-site inference.

• His judgment is very direct: almost everything with a chip may carry an AI model in the future.

VI. The real challenges of AI lie not only in model capabilities but also in security, identity, financial flow, organizational, and institutional resistance

• On security, his judgment is very sharp: almost all potential security bugs will be easier to discover, and a "computer security disaster" may occur in the short term.

• But he also believes that programming agents will scale the ability to patch vulnerabilities; in the future, the way to "protect software" may be to let bots scan and fix it.

• On the identity issue, he believes that "proof of bot" is not feasible because bots will become increasingly powerful; the truly viable direction is "proof of human," which is a combination of biometrics, cryptographic verification, and selective disclosure.

• He also discusses a frequently overlooked issue: if agents are to operate in the real world, they will ultimately need money, payment capabilities, and even some form of banking accounts, cards, or stablecoin-like infrastructure. On the organizational level, he borrows from the framework of managerial capitalism, believing that AI may reinforce founder-led companies because bots excel at reporting, coordination, documentation, and a large amount of "managerial work."

• However, he does not believe society will quickly and smoothly accept AI: he cites examples such as professional licenses, unions, dockworker strikes, government departments, K-12 education, and healthcare to illustrate that there are many institutional speed bumps in the real world.

• His judgment is that both AI utopians and doomsayers tend to overlook one point: just because technology is possible does not mean that 8 billion people will immediately change.