Research spotlights and articles on AI-driven software development from the CodeLab team at William & Mary. https://codelab.sh/blog.html en-us Tue, 08 Apr 2026 00:00:00 GMT amastro@wm.edu (Antonio Mastropaolo) amastro@wm.edu (Antonio Mastropaolo) https://codelab.sh/favicon.svg https://codelab.sh/blog.html https://codelab.sh/blog/secrets-in-the-synapses.html How generative AI is redefining steganography — hiding information not in what a model says, but in what the model is. amastro@wm.edu (Antonio Mastropaolo) Sun, 01 Mar 2026 00:00:00 GMT https://codelab.sh/blog/secrets-in-the-synapses.html https://codelab.sh/blog/mind-the-overlap.html How hidden data leakage from interdataset code duplication artificially inflates benchmark performance, and what we can do about it. amastro@wm.edu (Antonio Mastropaolo) Sun, 01 Feb 2026 00:00:00 GMT https://codelab.sh/blog/mind-the-overlap.html https://codelab.sh/blog/human-or-machine.html How AI-generated text challenges authenticity and trust in digital communication, and why transparency and human oversight remain essential. amastro@wm.edu (Antonio Mastropaolo) Thu, 01 Jan 2026 00:00:00 GMT https://codelab.sh/blog/human-or-machine.html https://codelab.sh/blog/prompt-alchemy.html Automating prompt refinement for code generation — how Prochemy closes the loop between prompt engineering and test-driven development. amastro@wm.edu (Antonio Mastropaolo) Mon, 01 Dec 2025 00:00:00 GMT https://codelab.sh/blog/prompt-alchemy.html https://codelab.sh/blog/breaking-bottlenecks.html How TightLLM uses adaptive offloading — trading compute for transfer and streaming weights in chunks — to achieve up to 23x throughput gains on modest hardware. amastro@wm.edu (Antonio Mastropaolo) Sat, 01 Nov 2025 00:00:00 GMT https://codelab.sh/blog/breaking-bottlenecks.html https://codelab.sh/blog/code-chaos-clarity.html Why scaling models isn't enough — how neurosymbolic software engineering combines neural learning, symbolic reasoning, and controlled chaos for trustworthy AI-assisted development. amastro@wm.edu (Antonio Mastropaolo) Wed, 01 Oct 2025 00:00:00 GMT https://codelab.sh/blog/code-chaos-clarity.html https://codelab.sh/blog/when-databases-age.html How continuous operation gradually degrades SQL Server and MySQL — and why proactive monitoring and software rejuvenation matter for long-term reliability. amastro@wm.edu (Antonio Mastropaolo) Mon, 01 Sep 2025 00:00:00 GMT https://codelab.sh/blog/when-databases-age.html https://codelab.sh/blog/smarter-not-harder.html How evolutionary algorithms can identify elite training samples — achieving up to 50% accuracy gains with 98% less energy by training on just 10% of the data. amastro@wm.edu (Antonio Mastropaolo) Fri, 01 Aug 2025 00:00:00 GMT https://codelab.sh/blog/smarter-not-harder.html https://codelab.sh/blog/pixels-of-deception.html How a single invisible noise pattern — evolved through genetic algorithms — can fool deep neural networks across thousands of images without any access to the model's internals. amastro@wm.edu (Antonio Mastropaolo) Tue, 01 Jul 2025 00:00:00 GMT https://codelab.sh/blog/pixels-of-deception.html https://codelab.sh/blog/ai-reshaping-lives.html Introducing LUNA — a universal analysis framework for making LLM behavior interpretable, addressing hallucinations, biases, and output inconsistencies. amastro@wm.edu (Antonio Mastropaolo) Sun, 01 Jun 2025 00:00:00 GMT https://codelab.sh/blog/ai-reshaping-lives.html