Open-weight AI: choosing control as well as a model
Proprietary APIs, downloadable weights, and hybrid architectures: choosing an AI model also involves vendor dependency, costs, and data management.
Notes from my work.
I write about development, automation and problems I come across in my work. I try to explain technical choices in a way that is useful even if you do not build software.
Articles
Ecommerce: trust breaks down between inventory and customer support
Buying a laptop reminded me how much real stock availability, delay updates, and customer support affect the customer experience.
Europe and AI: the risk of depending on others’ decisions
If access to a model changes, what happens to the products using it? A reflection on technological autonomy, expertise, and European alternatives.
Becoming a freelancer: starting at the right time
From a LinkedIn announcement to deciding to wait until January: personal products, new collaborations, and a master’s degree to prepare for launch.
When users go back to Excel, listen to the signal
A spreadsheet running alongside business software can reveal unnecessary steps, ignored exceptions, and real needs: understanding why matters more than banning it.
From graduation to independent work: the choice that started it
University, early projects, and working as a DJ: why I chose to pursue independent software development after my bachelor’s degree in Computer Science.
Chinese AI models: the ecosystem matters beyond benchmarks
DeepSeek, Qwen, and Kimi show why model distribution matters to developers: quality, infrastructure, licenses, and the ability to switch.
Before adding AI, stop making people act as APIs
Copying data between email, Excel, and business software is a process problem. Before introducing an AI agent, identify where the workflow breaks down.
Custom software or off-the-shelf: how an SME should choose
When custom software is worth it and when off-the-shelf business software is enough: practical criteria of cost, process, and risk to decide without waste.
AI APIs through proxies: risks and controls for informed business use
A cautious guide to unofficial proxies for AI models: data, service identity, continuity, quality, and the checks to make before integration.
Automating Excel-based processes: the method used for GigFlow
How to tell when a spreadsheet has become a bottleneck and turn the process into software without automating the chaos.
The value of software beyond the interface: what an SME should assess
Databases, rules, security, and maintenance: a practical guide to the less visible work that makes business software reliable.
The next projectStart with a conversation
Tell me about
your project.
Have an idea or something you would like to improve? Get in touch and we can work out together how I can help.
Prefer a call?