Standardization is key for any business that wants to maximize the impact of adopting Gen AI for software engineering, and code reviews are ripe for improvement. Here’s how to get started.
Ben Lloyd Pearson’s Post
More Relevant Posts
-
AI makes coding faster, but the heart of software development—solving business problems—hasn’t changed. Learn how to balance AI speed with human insight to create smarter, more robust solutions. Check out our latest blog at illuminai.select! https://zurl.co/G7CV
-
The delta between POC and Production with Agents is gigantic. Why is that? We are not treating our AI projects like we treat traditional software development. Our AI features need to have all the traditional software engineering best practices. - make files - ci/cd - tests And they need to have some extra oomph to account for the probabilistic nature of AI. Yes, that means evaluation.
-
Great work, OpenAI! SWE-bench Verified is a big leap in testing AI's ability to solve real-world software issues. A technical question: How diverse is the dataset in terms of verified authors, and what criteria were used to select these experts?
We're releasing a new iteration of SWE-bench, in collaboration with the original authors, to more reliably evaluate AI models on their ability to solve real-world software issues.
-
Software Modernisation | Product Development | Application Development | Codurance Head of Sales
Curious about the impact of AI on software engineering? The Codurance e-book, "Is AI about to revolutionise software development?" breaks down everything you need to know, from models and adoption to security. Download it today and prepare for the AI-driven future of development.
-
My take on this as well: Functions: logical units of code in software engineering. MapReduce: logical units of code for large scale data engineering. Agents: are the logical units of code in current state of AI / software + data engineering. The shift is that rather than have to be, because simulation-based (e.g. Monte Carlo) decisions and “probabilistic” algorithms are slowly gaining in popularity. In this paradigm, agents are actually better fit than deterministic output sometimes - since you can do 1000 runs to come up with something that mimics business simulation across many possibilities, given how certain the agent is with its answers and prior agent answers. Stack some agents on top of agents and you have a life simulator, as long as you have most agents loosely mapped right. Am I crazy or am I crazy.
It's getting clearer that AI is a superset of Software Engineering. Not the other way around
-
MIS || Data analysis & Developer || solving problems though the use of Data and Technology
eraser.io really took software design to whole new level especially with the new AI Diagram feature
-
-
What would you pay for an AI engineer that picks up tickets by itself and clears half your backlog in a few minutes? This is our actual codebase. It's quite complex and orchestrates our dev environment VMs. This ticket might have previously taken us a few hours end to end, but Engine has a PR to review in less than a minute. This is what AI-powered software engineering for real high-performing teams looks like.
