Tom Van Cutsem

Tom Van Cutsem

@tvcutsem.bsky.social

Computer scientist | Professor at KU Leuven | views are my own

103 followers84 following56 posts

I also appreciate that they give the audience the full context that Erich Gamma (vscode's creator) already made a name for himself long before this project as one of the co-authors of the Design Patterns book. That book teaches you the essence of OOP: how to design flexible, extensible software.

After a series of scripted scenes, students arrive in a free-play sandbox: they can click a node to crash it, drag between two nodes to partition them, and inject client commands.

Consensus is where my Distributed Systems students' eyes glaze over. They nod along to a Raft animation, then stare at slide 6 of pseudocode like it's a different subject. The translation between the two is the hard part — and I left it entirely to them. So I fixed that. 🧵

View full thread

3) Git saved the day: not only did its branching enable parallel development - I also hit my usage limits halfway (obviously - who doesn't?) but Claude was able to pick up the pieces of its sub-agents after inspecting partial work and git diffs.

4) In prep of this redesign I spent a good hour with Claude to ensure I could build the site locally - after having outsoured this to github-pages for the last ~8 years. Local builds alllowed the agent to 'close the loop' and iterate on its design faster.

View full thread

Lessons learned: 1) Jekyll (or any other static site generator) is a perfect match to modern AI tools: both content and style are accessible as plain text. Claude grokked the codebase - easily.

2) I did not just ask for 1 redesign, I asked for three, then picked the one I liked most but cherry-picked good features from the other two. I explicitly asked to develop each in a separate branch, so all three designs were developed in parallel and could be served simultaneously.

View full thread

The security model is compelling too. Projects like IronCurtain expose high-level APIs to agent code, so that when it makes calls like `gmail.sendEmail({to: "bob"})` instead of making raw HTTP requests, the host can enforce semantic policy. It's basically object capabilities, applied to AI.

View full thread

Cloudflare and Anthropic are converging on "Code Mode": instead of LLMs issuing tool calls one at a time, the LLM writes a TypeScript program against a typed host API, executed in a JS sandbox.

Why TypeScript? Not because TS is "better" than Python, but rather because of a quirk of JavaScript's history: the language has zero built-in powers. No file I/O, no network. Everything comes from the host environment. Browser → Node → and now: AI agent harnesses.

Today I was doing some research on standalone Javascript runtimes. The OG in this space is of course node.js. The two contenders for the 'modern alternative' to node seem to be Deno and Bun.

View full thread

Judging from Github activity, both engines seem to be under active development with a roughly equal-sized pool of contributors. Here's their 'star' history. Bun is the younger project but seems to be gaining traction faster.

Today I was doing some research on standalone Javascript runtimes. The OG in this space is of course node.js. The two contenders for the 'modern alternative' to node seem to be Deno and Bun.

View full thread

Both Deno and Bun support TypeScript source files natively, and they both integrate common devtools such as build system, package manager and test runner into the core.

Today I was doing some research on standalone Javascript runtimes. The OG in this space is of course node.js. The two contenders for the 'modern alternative' to node seem to be Deno and Bun.

IT Industry professional in Europe? Come to Leuven, Belgium this summer, join the SecAppDev course to plug your brain into everything application security, and you won't regret that you did. This year, a special focus on (how could it not be) AI security.

SecAppDev@secappdev.org

SecAppDev 2026 is taking shape 🚀 We’ve just released the first program details, with initial speakers and workshops now live. Expect deep dives into application security, modern threats, and hands-on training. Full details on secappdev.org #appsec #infosec

AI is a huge force multiplier, but only if you have fundamentals to steer it and review it well. That’s exactly why CS education still matters (arguably more than ever). #AI #VibeCoding #CSedu #SoftwareEngineering

View full thread

Steve Jobs once famously called computers "bicycles for the mind". The last few days I've been coding and reviewing math problems with the help of recent LLMs, and the clarity with which they lay out coding & math problems is really staggering.

So, despite all the flaws and limitations of LLM use in general, for math/coding, I couldn't help but be reminded of Jobs's quote. What do you call a rickety bicycle that usually gets you to where you want to be, but sometimes breaks down unexpectedly? Velocipedes for the mind? :-)

And no, he's not hyping AI or making silly predictions such as "nobody will need to write code anymore". It's a nuanced opinion from somebody who saw the power of deep learning coming long before most others did.