Props to @shanselman for writing about the elephant in the room in big tech.
-
Props to @shanselman for writing about the elephant in the room in big tech.
AI coding tools increase senior engineer productivity while creating an “AI drag” for junior engineers who lack judgment or context to validate AI outputs.
Companies optimizing for the near-term output by reducing junior hiring, are undermining the future supply of experienced engineers. We likely need to move to apprenticeship like models in the future to give juniors experience.
https://dl.acm.org/doi/10.1145/3779312@carnage4life @shanselman I have long thought apprenticeship is the ideal model for the software industry.
A large part of what we do is not easily taught but acquired through experience and while that can be acquired on the job it can be hard to be exposed to without the right opportunities.
What we now see with LLMs is a prime opportunity for our roles as senior / architect level engineers to step into a role where we direct machines and educate the next generation to take our roles.
-
Props to @shanselman for writing about the elephant in the room in big tech.
AI coding tools increase senior engineer productivity while creating an “AI drag” for junior engineers who lack judgment or context to validate AI outputs.
Companies optimizing for the near-term output by reducing junior hiring, are undermining the future supply of experienced engineers. We likely need to move to apprenticeship like models in the future to give juniors experience.
https://dl.acm.org/doi/10.1145/3779312@carnage4life @shanselman instead of seniors spending their time coercing the desired outputs from an LLM, they could be reviewing the code from juniors instead, and helping them learn, something which is seemingly impossible for this generation of LLMs
-
Props to @shanselman for writing about the elephant in the room in big tech.
AI coding tools increase senior engineer productivity while creating an “AI drag” for junior engineers who lack judgment or context to validate AI outputs.
Companies optimizing for the near-term output by reducing junior hiring, are undermining the future supply of experienced engineers. We likely need to move to apprenticeship like models in the future to give juniors experience.
https://dl.acm.org/doi/10.1145/3779312@carnage4life @shanselman We even have somwhat of a precedent for that: during the dotcom crash, a lot of CS students dropped out... and that resulted in at least a decade of hiring struggles (and possibly in inflated salaries as a result).
-
@carnage4life @shanselman I have long thought apprenticeship is the ideal model for the software industry.
A large part of what we do is not easily taught but acquired through experience and while that can be acquired on the job it can be hard to be exposed to without the right opportunities.
What we now see with LLMs is a prime opportunity for our roles as senior / architect level engineers to step into a role where we direct machines and educate the next generation to take our roles.
@carnage4life @shanselman teaching is the best possible way to find gaps in knowledge and become better at describing our roles.
So this becomes a symbiotic relationship.
We teach the next generation to think, to orchestrate, to architect and understand the machines we program. And the machines get the work done to write the code while humans do the hard part, that the software industry has pretended isn't the job for so long - defining the work and applying it to actual problems and customers
-
Props to @shanselman for writing about the elephant in the room in big tech.
AI coding tools increase senior engineer productivity while creating an “AI drag” for junior engineers who lack judgment or context to validate AI outputs.
Companies optimizing for the near-term output by reducing junior hiring, are undermining the future supply of experienced engineers. We likely need to move to apprenticeship like models in the future to give juniors experience.
https://dl.acm.org/doi/10.1145/3779312@carnage4life @shanselman And unlike junior devs, the AI never learns or improves; you're at their maximum right now.
-
Props to @shanselman for writing about the elephant in the room in big tech.
AI coding tools increase senior engineer productivity while creating an “AI drag” for junior engineers who lack judgment or context to validate AI outputs.
Companies optimizing for the near-term output by reducing junior hiring, are undermining the future supply of experienced engineers. We likely need to move to apprenticeship like models in the future to give juniors experience.
https://dl.acm.org/doi/10.1145/3779312@carnage4life @shanselman I think this is characteristic of anyone using AI: it's impossible to know if it's telling you the truth unless you're something of a subject matter expert, at which point, you may not really need its help.
-
Props to @shanselman for writing about the elephant in the room in big tech.
AI coding tools increase senior engineer productivity while creating an “AI drag” for junior engineers who lack judgment or context to validate AI outputs.
Companies optimizing for the near-term output by reducing junior hiring, are undermining the future supply of experienced engineers. We likely need to move to apprenticeship like models in the future to give juniors experience.
https://dl.acm.org/doi/10.1145/3779312@carnage4life looking at industries where apprenticeships have traditionally been how people were trained (eg trades) does not bode well for an apprenticeship model for software development.
-
@carnage4life looking at industries where apprenticeships have traditionally been how people were trained (eg trades) does not bode well for an apprenticeship model for software development.
@womble @carnage4life in what sense, specifically? https://blog.codinghorror.com/software-apprenticeship/
-
@womble @carnage4life in what sense, specifically? https://blog.codinghorror.com/software-apprenticeship/
@codinghorror not enough existing practitioners are willing to take on apprentices, for starters. There is projected to be a deficit of 300,000 skilled tradespeople by 2027 in Australia alone (https://www.abc.net.au/news/2025-11-13/construction-workforce-needs-to-double/106002738). A training model that doesn't produce enough workers isn't a good one, regardless of any other benefits it might have.
-
@codinghorror not enough existing practitioners are willing to take on apprentices, for starters. There is projected to be a deficit of 300,000 skilled tradespeople by 2027 in Australia alone (https://www.abc.net.au/news/2025-11-13/construction-workforce-needs-to-double/106002738). A training model that doesn't produce enough workers isn't a good one, regardless of any other benefits it might have.
@womble fair, probably worth blogging in detail as it relates to software dev
-
R relay@relay.infosec.exchange shared this topic
