Lightnews — Scholar-powered news

Reposted by Jo Van Eyck

Tim Schraepen @sch3lp.bsky.social · 1d

That’s a wrap for day one of #SocratesBE
Here’s a FOMO inducing photo of today’s marketplace.
What you don’t see is Vincent’s legendary lightning talk on Limburgse Vlaai. That happened Thursday evening already.

1 4

Jo Van Eyck @jovaneyck.bsky.social · 2d

Great idea!
Not serena-related, but I've been working on opening up my fsi.exe sessions over mcp and it has been very fun programming this way: github.com/jovaneyck/fs...

GitHub - jovaneyck/fsi-mcp-server: A wrapper around F# interactive (fsi) that exposes an MCP server endpoint for AI agents allowing for fsi IO.

A wrapper around F# interactive (fsi) that exposes an MCP server endpoint for AI agents allowing for fsi IO. - jovaneyck/fsi-mcp-server

github.com

1 1 3

Reposted by Jo Van Eyck

Don Syme @dsyme.bsky.social · 2d

F# folk - I've started a PR to add F# support to the Serena MCP server, that effectively allows the use of the F# language server from coding agents.

github.com/oraios/seren...

Please be welcome to contribute to get this over the line.

cc @chethusk.bsky.social

Thanks
Don

Add F# support using FsAutoComplete LSP by dsyme · Pull Request #662 · oraios/serena

This adds F# support by the usual path, using the standard F# LSP Using Serena with F# Projects This guide demonstrates how to use Serena's tools with F# projects through both the MCP protocol...

github.com

1 3 11

Reposted by Jo Van Eyck

Simon Willison @simonwillison.net · 7d

I've become one of those people who run multiple coding agents at once - here are my notes on embracing the parallel coding agent lifestyle simonwillison.net/2025/Oct/5/p...

Embracing the parallel coding agent lifestyle

For a while now I’ve been hearing from engineers who run multiple coding agents at once—firing up several Claude Code or Codex CLI instances at the same time, sometimes in …

simonwillison.net

25 14 160

Reposted by Jo Van Eyck

Gergely Orosz @gergely.pragmaticengineer.com · 11d

AI agents being able to run unit tests is SUCH a massive unlock.

If you're a dev who has utilized unit tests heavily: using a tool like Claude Code makes so much sense.

And if you don't yet use tests: with AI agents, don't see how you would not do it sooner or later...

8 5 61

Jo Van Eyck @jovaneyck.bsky.social · 23d

Autonomous development of a backend feature in a typical "enterprise app". No spec driven bullshit, no agent swarm madness and very acceptable results for my tastes:

github.com/jovaneyck/eS...

github.com

Reposted by Jo Van Eyck

Simon Willison @simonwillison.net · 23d

I'm ready to accept a definition of "agent" that I think is widely-enough agreed upon to be useful, at least in the AI engineering space:

An LLM agent runs tools in a loop to achieve a goal

(I've been dismissing the term as hopelessly ambiguous for years) simonwillison.net/2025/Sep/18/...

I think “agent” may finally have a widely enough agreed upon definition to be useful jargon now

I’ve noticed something interesting over the past few weeks: I’ve started using the term “agent” in conversations where I don’t feel the need to then define it, roll my eyes …

simonwillison.net

8 18 130

Reposted by Jo Van Eyck

Rob Bowley @robbowley.net · Sep 8

I don’t think people yet fully appreciate how the “lethal trifecta” restricts agentic AI use cases. When private data access, untrusted inputs, and external actions converge, you create a structural vulnerability we don’t yet know how to contain.

simonwillison.net/2025/Jun/16/...

The lethal trifecta for AI agents: private data, untrusted content, and external communication

If you are a user of LLM systems that use tools (you can call them “AI agents” if you like) it is critically important that you understand the risk of …

simonwillison.net

2 5 18

Jo Van Eyck @jovaneyck.bsky.social · Sep 1

Thanks for sharing your journey, it's been a highlight of my online reading summer!

1

Jo Van Eyck @jovaneyck.bsky.social · Aug 31

Ever wondered how AI coding agents like Claude Code, Github Copilot and Codex work? So did I! So of course I had to go and build one myself 😅

youtu.be/7oTPNr9APGE

#claudecode #codex #cursor #copilot #anthropic #openai #llm #agenticai #ai

I Built an AI Coding Agent in 200 Lines of Code.

YouTube video by Jo Van Eyck

youtu.be

1 2

Jo Van Eyck @jovaneyck.bsky.social · Aug 24

Can AI coding agents do Test-Driven Development (TDD) properly? They sure can!

In this week's video I take a look at two approaches: prompt engineering & using the tdd-guard library

📺 youtu.be/IVdYaVKuekk

#ai #aiassistedcoding #claudecode #tdd #testdrivendevelopment #tddguard

Can AI coding agents do Test-Driven Development (TDD)?

YouTube video by Jo Van Eyck

youtu.be

Reposted by Jo Van Eyck

Bits @bits.onethingwell.dev · Aug 18

LLMs and coding agents are a security nightmare

garymarcus.substack.com/p/llms-codin...

#ai #genai #llm #programming

LLMs + Coding Agents = Security Nightmare

Things are about to get wild

garymarcus.substack.com

4 5

Jo Van Eyck @jovaneyck.bsky.social · Aug 17

AI coding assistants promise 10x productivity gains. The research tells a different story.

Full breakdown here: youtu.be/MLSs8Asmthk

#aiassistedcoding #claudecode #githubcopilot #dora

Busting the AI coding productivity myth

YouTube video by Jo Van Eyck

youtu.be

1

Reposted by Jo Van Eyck

Amber Williams @notnotamber.bsky.social · Aug 15

AI may not make you a 10x engineer, but understanding how to use AI assisted coding frameworks can cost 20x less.

Diagram shows an experiment by @jovaneyck.bsky.social using MCPs with Claude Code vs without MCPs

1 1

Reposted by Jo Van Eyck

Mathias Verraes @mathiasverraes.bsky.social · Aug 15

It's the 10th anniversary of this joke. I originally posted it on Twitter, and I felt it deserved a new home. It has been quoted in books, presentations, academic lectures, and t-shirts, and it's probably the only thing I'll be remembered for 🤣

A selfie with a a gentleman I met at a conference in Denver CO, who made a t-shirt with my distsys joke printed on it.

1 2 11

Jo Van Eyck @jovaneyck.bsky.social · Aug 10

Everyone’s chasing 10x boost from coding assistants. In my coding I get nowhere near that promised 10x. But in other areas it comes close!

Inside my Obsidian vault, I built:

* Semantic search
* Interactive quizmaster
* slide decks to text

Watch here: youtu.be/nSh2BYJ29kY

#claudecode #obsidian

This combo makes smart note systems magical.

YouTube video by Jo Van Eyck

youtu.be

1 2

Reposted by Jo Van Eyck

Katie Mack @astrokatie.com · Aug 9

"PhD-level experts in your back pocket" is a completely nonsensical description of AI but a pretty good description of social media if you follow the right people

140 2K 10K

Jo Van Eyck @jovaneyck.bsky.social · Aug 6

But but what about all that extra RAM you need to run LLM's on edge devices? *ducks*

Jo Van Eyck @jovaneyck.bsky.social · Aug 2

AI coding assistants not working for you because your legacy codebase is too big? There's a solution!

In this video I share one-click and FREE solutions that give Claude Code a staggering 30x performance boost when working in large codebases.

youtu.be/UqfxuQKuMo8

#aiassistedcoding #claudecode

AI coding agents are useless on large codebases. Unless you do THIS.

YouTube video by Jo Van Eyck

youtu.be

Jo Van Eyck @jovaneyck.bsky.social · Jul 31

Are we talking about an accidental consequence of tab-monkeys or a conscious strategy to "make it easier" for LLM's?

1

Reposted by Jo Van Eyck

gabby @fullmoon.id · Jul 21

Along the same lines:

Comic strip:

Coworker A (drinking coffee): "Some day we won't even need coders any more. We'll be able to just write the specification and the program will write itself"

Coworker B (coding at a laptop): Oh wow, you're right! We'll be able to write a comprehensive and precise spec and bam, we won't need programmers any more!

And do you know the industry term for a project specification that is comprehensive and precise enough to generate a program?

Coworker A: Uh… no…

Coworker B: Code. It's called code

1 6 28

Jo Van Eyck @jovaneyck.bsky.social · Jul 26

Every software engineer needs to be taking a serious look at what AI coding agents will mean for them. In this video we'll go over the essentials:

* What to learn
* Who to follow
* Which tools to pick and why
* How to get started

youtu.be/xE3EGy_FAnQ
#claudecode #vibecoding

Getting started with AI coding agents

YouTube video by Jo Van Eyck

youtu.be

1

Jo Van Eyck @jovaneyck.bsky.social · Jul 24

Hands on the wheel: live-blogging a claude code programming session

This post is an unvarnished look at what it's like to work with an AI assistant in real time. Not a curated demo. Not a sanitized GitHub Copilot ad. This is what actually happened when I paired with Claude code to work on a…

Hands on the wheel: live-blogging a claude code programming session

This post is an unvarnished look at what it's like to work with an AI assistant in real time. Not a curated demo. Not a sanitized GitHub Copilot ad. This is what actually happened when I paired with Claude code to work on a non-trivial. I wanted to explore this area a bit further; because "hammering out another CRUD endpoint" is something I've seen Claude do quite successfully. I wanted to push things a bit further. And yes; I kept my hands somewhere near the steering wheel the entire time.

jvaneyck.wordpress.com

1

Jo Van Eyck @jovaneyck.bsky.social · Jul 21

This has to be one of the most thoughtful, reasonable yet realistically forward looking stances I've come across. Thanks for sharing!

Jo Van Eyck @jovaneyck.bsky.social · Jul 21

youtu.be/2sLT2gugnIY

Lots of developers are treating LLMs like a slot machine.

Effective devs need to learn how to stack the odds in their favor. We'll go over some metaphors to help you break away from gambling and towards effective coding leverage. 🎰🧞🤖🐶🖥️
#llms #claudecode #aiassistedprogramming

Dear devs, please Stop using AI like a slot machine

YouTube video by Jo Van Eyck

youtu.be

2