AI Coding Assistants – Something No One is Talking About

Last week the Traffyk team had the pleasure of attending a panel organised by Konnecta “Engineering in the Age of AI”. What was surprising was that majority of questions were focused on trust: “How can I get clients to trust [gen]AI?”, “how can I trust [gen]AI?”. Commonwealth Bank Chief AI Engineer Blair Hudson  made a good point that he doesn’t trust ChatGPT but he trusts his judgement in interpreting its output.

Regarding the issue of ‘lack of trust’, I think there are two things at play here:

1.       Overpromise by the AI leaders (OpenAI, Anthropic, NVIDIA)

2.      General lack of understanding and AI literacy in the population

The second point can be forgiven. However, the first point is irresponsible as some of the worse consequences are job loss (then subsequent backtracking when AI cannot replace a human), and even death when children are coached to commit suicide by AI.

What is the solution? I think it’s now on the AI experts to correct the narrative, educate and set expectations. Reejig CTAO Dr. Shujia Zhang shared very grounded thoughts on how AI should be adopted in a collaborative process with humans and should enable us to do more of the thinking.

The most progressed use-case where AI/human collaborate is coding assistants. Roman Dydiuk, our Sr Full-Stack Engineer, sparked an interesting discussion. With the current rapid adoption of AI coding assistants, use of Stack Overflow (reddit for coders) has dropped off a cliff. This is because the AI is trained off Stack Overflow so there’s no longer a need to ask questions on a forum and wait for a human answer. But if human code developers’ role is now primarily focused on creativity, for coding assistants to progress, i.e., new packages, languages, solutions (creativity) to be harnessed, the Stack Overflow training loop is no longer supported. Ideas are no longer shared on a public forum and instead shared human-to-AI (company).

The implications of this are:

1.        New AI models will not have a wealth of up-to-date information on Stack Overflow to train from. This will make open-source coding assistants less competitive in the future.

2.        Creativity in coding may plateau/remain siloed.

3.        There will be a reliance on the big AI companies for coding assistants as they are the ones with the means to keep their AI’s coding knowledge current.

4.        The value of creative software engineers will continue to skyrocket (see Andrew Tulloch's latest career move)

At the end of the day, I don’t think the AI leaders will stop the hype. To mitigate this, AI experts need to cut through the hype and establish trust with the community. For people working with genAI, your greatest asset is your decision making and creativity.

Remain critical of genAI outputs and don't stop the human-human collaboration.

‍

13 October 2025