The AI dyad

Working in collaboration with a model, in isolation from your team or cross-functional collaboraters —> the AI dyad.

Might not be for everyone. Part of the pleasure and incentive to work for many people is the interaction with other people, whether collaborating, teaching, learning from, competing, completing, validating. Stripping that relational richness for a barebone robotic life will be a loss for many. There will be a subset who do not mind the replacement of their daily interactions with people to become with robots.

The end of the team

In a context of atomized workers, each augmented and interacting with AI rather than people. "Jusk ask model and it will tell you". Breaking of communication. I don't need to ask you if I can ask model. Conversely, why are you asking me, you can just ask model.

≥ Flat team: 1 manager, 50 engineers. One engineer's output replaces that of a whole team. Each engineer becomes a team, or a CEO of agents, or its own entity.

Self-reliance and the complete autonomy myth

For who views reliance and working with others as a drag, a necessary price to pay to overcome their own limitations, the AI dyad can feel like liberation. I can finally do what I want, leveraging the AI rather than others to both complement me (in the things I don't know how to do) and amplify my outcome.

Augmentation, i.e bringing in complementary knowledge in my dyad.

                    ---> the AI dyad feeds itself 
                    |
2 divergent paths --- 
                    |  
                    ---> the cross-functional team path

Lack of involvement from others during the work. Difficulty in reviewing, contributing or giving feedback post-fact.

Others lose context on my work, they have higher difficulties in understanding it, reviewing it. -> So the cycle of them being excluded from it.

1. work in dyad
2. ask questions within dyad
3. get feedback from dyad
4. output

No other person involved in the process heightens the barrier to engage with it when it is completed.

Death of collaboration

The dream of augmenting your thought process while you are within your dyad. You reproduce your collaborator virtually through the AI to never leave your dyad.

• Loss of texture due to the narrow context the AI has, and that you have to enrich either through constant transmission/tweaking (make a peanut butter sandwich instructions) or laboriously make available to it
• Loss of viewpoint diversity
• Loss of deliberation: you're the only one deciding, or sometimes only the AI does. Noone else weighs in on the design, or overall approach.

Is there a third path, apart from the cross-functional team path without AI, and the atomized AI dyads path?

Signs of the AI dyad path

• Large bodies of work produced in isolation
• Strong review burden: no context, no translation or transmission, large volume of output
• Multiple instances of augmentation: rather than talking with cross-functional peers, let me emulate them virtually.
• Post-fact: no involvement of others, no deliberation, creation of something and then asking 'use it and see' or 'how to make it work'

Bottlenecks:

On the AI dyad path, the bottlenecks quickly appear:

• 'Now review it!' -> but no context, too much volume
• 'Now use it!' -> but what was it scoped for?
• 'Now integrate it into production/the system' -> but what were the requirements, what are best practices