Person
1 episodes · 17 quoted claims
Tim Sears says that before AI, requirements gathering was a lengthy process done carefully so that software engineers could focus on building exactly the right thing.
“requirements gathering, which is a pretty lengthy process and it had to be done super carefully so that your precious software engineers could focus on doing exactly the right thing”
Tim Sears names the large improvement in foundation model quality over about three years as the top trend he expects to continue.
“So the number one thing is that the quality of the foundation models has improved a lot.”
Tim Sears says AI-generated code can be checked with compilers, tests and deployment in test or production environments.
“You have compilers and you have you have tests and you have, you know, you can see how it works in deployment, whether it's in test or production”
Tim Sears argues that code's built-in validation methods allow large language model output to be checked with less human judgment, lowering cost.
“So you have ways of validating the output of LLMs with less human judgment, which translates into lower cost.”
Tim Sears predicts that sprints in software development will fade into essentially continuous iteration as AI becomes a catalyst for teamwork.
“So that means sprints, things like that, if you know that word, are are gonna fade into essentially continuous iteration.”
Tim Sears says AI productivity gains tied to individuals are uneven, while gains tied to teams will bring big benefits.
“So, when productivity gains are tied to teams, we're gonna see big benefits. Until then, tied to individuals, we see some, but progress will be very uneven.”
Tim Sears expects long-term improvements in AI inference speed and, more importantly, inference cost.
“But I expect the longer term is we're going to see a lot of improvements in inference speed and inference costs, more importantly.”
Tim Sears says that cutting about two thirds of a call center's costs would turn it into a hybrid of humans and computer programs.
“But if we want to save some costs and maybe take two thirds of the costs out of a call center, now it's a hybrid between humans and a bunch of computer programs”
Tim Sears says AI gives leverage that lets small teams do what large teams once did.
“In fact, I think the leverage provided by AI is that, you know, small teams can do what big teams used to do.”
Tim Sears says HTEC's client partners, effectively its salespeople, can record calls and turn the conversations into requirements, starting on the initial sales call.
“our client partners, effectively our salespeople, could be recording calls and turning those things, those conversations into requirements.”
Tim Sears says the developer role is becoming more elevated, focused on getting architecture right and handling security early.
“the developer job is is le- is more elevated in the sense of making sure the architecture's right, making sure that security matters are being dealt with early”
Tim Sears says people should take ownership of AI-written documents they send rather than attributing them to ChatGPT.
“if somebody sends me a document written by AI, I'd like them to own it and not tell me that ChatGPT just wrote it for them.”
Tim Sears says private equity firms are mandating more AI use in their portfolio companies, creating top-down pressure for change.
“we see essentially owners of companies like private equity firms, for example, mandating more AI in their portfolio companies.”
Tim Sears disputes the claim that SaaS is dead, saying enterprise software is changing but not going away.
“you hear things like SaaS is dead and so forth. I think business models will change, the way we work will change, but enterprise software is changing, but it's not going away.”
Tim Sears says that for twenty to thirty years businesses have adapted their ways of working to fit existing software platforms.
“for the past, you know, twenty, thirty years, we've been benefiting from these great platforms, but in many ways, we've had to mold our, uh, ways of working to existing platforms.”
Tim Sears argues the traditional buy-versus-build decision is becoming a build-versus-build choice because all options involve building.
“It's a really a build versus build. You're going to have an opportunity to build it this way or build it that way”
Tim Sears predicts that within a few years, near-perfect enterprise software will be the average expectation.
“But in a few years, perfection in enterprise software is going to be average.”