𝗦𝗼𝗳𝘁𝘄𝗮𝗿𝗲 𝗘𝗻𝗴𝗶𝗻𝗲𝗲𝗿𝗶𝗻𝗴 𝗪𝗶𝗹𝗹 𝗕𝗲 𝗔𝘂𝘁𝗼𝗺𝗮𝘁𝗮𝗯𝗹𝗲 𝗶𝗻 𝟭𝟮 𝗠𝗼𝗻𝘁𝗵𝘀

This is what Anthropic CEO Dario Amodei said at the World Economic Forum in Davos last week. His exact words: “I think we might be 6 to 12 months away from when the model is doing most, maybe all of what SWEs do end-to-end.”

It is a bold prediction, but the headlines omit the part about others' views.

Demis Hassabis, Nobel Prize-winning CEO of Google DeepMind, said current AI is “𝘯𝘰𝘸𝘩𝘦𝘳𝘦 𝘯𝘦𝘢𝘳” human-level intelligence. He gives AGI a 50% chance within the decade and says we need “one or two more breakthroughs” beyond existing technology.

Yann LeCun, Turing Award winner and former Meta AI chief, went further. He called the industry “𝘤𝘰𝘮𝘱𝘭𝘦𝘵𝘦𝘭𝘺 𝘓𝘓𝘔-𝘱𝘪𝘭𝘭𝘦𝘥” and stated that large language models will never achieve human-level intelligence.

Three of the most respected minds in AI. Fundamentally incompatible views. Same stage.

The headlines also skip what Amodei actually described. His engineers “𝘥𝘰𝘯’𝘵 𝘸𝘳𝘪𝘵𝘦 𝘤𝘰𝘥𝘦 𝘢𝘯𝘺𝘮𝘰𝘳𝘦.” They let the model write it. They edit it. They do “𝘵𝘩𝘦 𝘵𝘩𝘪𝘯𝘨𝘴 𝘢𝘳𝘰𝘶𝘯𝘥 𝘪𝘵.” But this means, obviously, that it is not a replacement. That is role transformation. The engineer is still there, and important.

We have seen this pattern before. Self-driving cars were supposed to be here by 2018. Level 5 autonomy was “𝘫𝘶𝘴𝘵 𝘢𝘳𝘰𝘶𝘯𝘥 𝘵𝘩𝘦 𝘤𝘰𝘳𝘯𝘦𝘳.” Tesla still requires hands on the wheel. LeCun put it like this: 𝗶𝗺𝗽𝗿𝗲𝘀𝘀𝗶𝘃𝗲 𝗱𝗲𝗺𝗼𝘀 𝗱𝗼 𝗻𝗼𝘁 𝗲𝗾𝘂𝗮𝗹 𝗱𝗲𝗽𝗲𝗻𝗱𝗮𝗯𝗹𝗲 𝗱𝗲𝗽𝗹𝗼𝘆𝗺𝗲𝗻𝘁. There is a considerable distance between a demo and a system you put in people’s hands daily.

But let's consider who benefits from this story. Amodei runs a company valued in the billions, competing for talent, funding, and market standing against OpenAI and Google. Bold predictions attract attention, investment, and impetus to adopt their products. This is not a conspiracy. This is an incentive structure. 𝗪𝗵𝗲𝗻 𝘀𝗼𝗺𝗲𝗼𝗻𝗲 𝘁𝗲𝗹𝗹𝘀 𝘆𝗼𝘂 𝘁𝗵𝗲𝗶𝗿 𝗽𝗿𝗼𝗱𝘂𝗰𝘁 𝘄𝗶𝗹𝗹 𝗰𝗵𝗮𝗻𝗴𝗲 𝗲𝘃𝗲𝗿𝘆𝘁𝗵𝗶𝗻𝗴, 𝗮𝘀𝗸 𝘄𝗵𝗮𝘁 𝘁𝗵𝗲𝘆 𝗮𝗿𝗲 𝘀𝗲𝗹𝗹𝗶𝗻𝗴.

The fundamental transformation is already here. AI coding tools make developers more productive. Anthropic’s Cowork product was built in a week and a half, almost entirely with their AI tools. But someone still decided what to build. Someone evaluated if it worked. Someone made architectural decisions. Someone owned the outcome.

𝗧𝗵𝗲 𝘀𝗵𝗶𝗳𝘁 𝗶𝘀 𝗳𝗿𝗼𝗺 𝘁𝘆𝗽𝗶𝗻𝗴 𝗰𝗼𝗱𝗲 𝘁𝗼 𝗱𝗶𝗿𝗲𝗰𝘁𝗶𝗻𝗴 𝘀𝘆𝘀𝘁𝗲𝗺𝘀 𝘁𝗵𝗮𝘁 𝘁𝘆𝗽𝗲 𝗰𝗼𝗱𝗲. That requires different skills. No skills.

The question is not whether AI will change software engineering. It already has. The question is whether, 12 months from now, we will look back on this prediction the way we look at 2018's self-driving promises.

What do you think, is this a transformation or a hype cycle?