In many organisations, AI pilots don’t fail. They simply stop moving. The pilot runs successfully, results are shared internally, and the initiative is labelled a success. Yet months later, the AI solution is still not part of day-to-day operations. Teams continue working the same way they always have, and the...
Best of Nibbles 2025
A Curated Collection of Articles That Will Stand the Test of Time Every month, Nibbles attempts something simple and deceptively hard:to curate articles that are worth your time. Not trending.Not loud.Not driven by the algorithm of the week. As the year comes to a close, we found ourselves asking a...
The Automation Trap: Why Most Companies Automate the Wrong Things First
A few months into an automation initiative, the same question starts circulating quietly inside organisations:“Why are we automating so much, yet seeing so little change?” Dashboards look better. Tools are in place. But workflows still stall, teams still intervene manually, and exceptions still pile up. The promise of automation feels...
Inside GitHub’s New Code Search Engine
GitHub has rolled out major improvements to its search backend — rewritten in Rust with Zing indices and near-instant streaming search across massive repos. This upgrade dramatically improves cross-repo discovery, regex support, and monorepo developer experience. Read more: https://github.blog/engineering/architecture-optimization/the-technology-behind-githubs-new-code-search/ What is the future of platform engineering? Focus on the ‘Why’...
What Went Wrong with the AWS Outage
The massive outage in Amazon Web Services’ US-EAST-1 region was traced to a race condition in the DNS automation of the DynamoDB service. An empty DNS record and automation failures propagated across multiple AWS services, disrupting dozens of major websites and apps. Read more: Network World article Engineering Leaders on...
What I’ve Learned from Talking to Top Engineering Leaders in 2025
Engineering execs share insights on AI adoption, scaling distributed teams, and the tradeoffs of speed vs stability. If you want to read more, click here. What Engineering Managers Need to Know for 2025 Shifting expectations for managers: agentic systems, memory-aware AI tools, and new skills for team leadership.If you want...
GPT-5 for Developers —
OpenAI’s newest API models (gpt-5, mini, nano) focus on stronger coding and agentic tool-use with clear migration notes.Read more: https://openai.com/index/introducing-gpt-5-for-developers/OpenAI Realtime API (gpt-realtime) GA launch adds better speech-to-speech, SIP calling, image inputs, and remote MCP servers for production voice agents.Read more: https://openai.com/index/introducing-gpt-realtime/OpenAI LLM Evaluation at Booking.com A practical playbook: strong-model...
How I keep up with AI progress (and why you must too)
"How I Keep Up With AI Progress" outlines a deliberate strategy to stay informed amid the noise of AI hype and skepticism. The author advocates curating a high-signal feed from credible voices (like Simon Willison, Karpathy, and select researchers), following original sources from AI labs, and practicing intentional daily skimming...
Key take‑aways from Andrej Karpathy’s keynote “Software Is Changing (Again)”
Three paradigms of software Software 1.0 – traditional hand‑written code.Software 2.0 – neural‑network “weights as code”, produced by training on data.Software 3.0 – large‑language‑model computers that you “program” with English prompts; the prompt is now the source code. Take a look at the video here Learnings from two years of using AI tools for software engineering...
Nibbles 50
For our 50th edition we wanted to go back and shortlist five articles that stand the test of time, in other words classical. In other words if one has to read only 5 articles of the 200 odd curation, we are confident that these 5 will stand the test of...