Totally. My agents built a Linux From Scratch system and have maintained it for over two months: daily system and security updates, occasional kernel rebuilds, and everything compiled from source. No package manager.
https://t.co/7l5diKc5f2
@Spotify. @Klarna. @Lovable. @WeAreLegora. Why do so many breakout tech companies come out of Stockholm and should you build your next startup here?
To answer these questions, I spent time with the people shaping this ecosystem:
@ElisHodzic02 from https://t.co/9vIm6AB4tF → building one of Stockholm’s most talent-dense founder communities
@felixhhaas from @Lovable shaping the design of the fastest-growing startups in the world
Hanna Haas from @norrsken_org → helping make Stockholm one of Europe’s strongest hubs for impact entrepreneurship
Stephen Townsend from https://t.co/GBG6kDPV4E → building a health tech startup for long covid and post-viral care
@zorankokolanski & Olivia from @joinfivedegrees → bringing together some of Europe’s hottest founders and talent
Episode 1 is live.
Link in comments :)
Built a keyboard-first breadcrumb navigator for GitHub.
Press Ctrl+B to jump across repos, folders, and files without breaking flow.
https://t.co/7mwKcvqJ0y
How do you see interfaces evolving, and do you think GraphQL becomes the default as systems get more AI-native?
And if interfaces do stay defined, where do you expect companies to re-implement systems purely to gain interface control? Which domains or processes come to mind?
Software is beginning to feel like orienteering: navigating unfamiliar terrain while still reaching every required checkpoint (security, audit, permissions) before crossing the finish line.
Will interfaces evolve as well and enable fully fluid, end-to-end business processes?
- Interfaces stay defined (for security, governance, reliability), and companies will increasingly need to build or own more of their stack to control interfaces for high-stakes end-to-end workflows.
A couple of scenarios feel realistic to me:
- Interfaces become fluid too. Instead of predefined REST endpoints that both standardise and limit what you can do, we get more “shapeless” access to a system’s capabilities. If that happens, GraphQL feels more future-proof than REST.
What I keep coming back to is the next step: moving beyond software boundaries and making business processes themselves fluid. That will take more than better models. It will require real control over the interfaces across your system landscape.
It feels a lot like orienteering (https://t.co/YGetW4RJvW
): you need to reach the finish line through unknown terrain, but you still have to pass specific checkpoints along the way.
I used to think in terms of defined operations and workflows. Now I am leaning towards a model where you mainly define checkpoints (regulatory, security, audit, permissions) and keep the business logic itself fluid, meaning it is generated and adapted to the customer’s context.
Excited for AI Tinkerer London tomorrow at @monzo !!
A real mix of practical and experimental talks...
- Task Specific LLMs & LoRA - Dan Cooper @ monzo
- Stable Diffusion As Game Engine Renderer - @thomastraum
- Finetuning Mistral 7B with Synthetic Data - @felixbrockm
- Falcon OS, LLM Operating System - @HeikoHotz
- Finetuning Mistral 7B on Coding Tasks - @skcd42
- The Dangers of Vanilla RAG - @neal_lathia
Join me at the AI Tinkerers London January Meetup next week organised by our friends at @bloopdotai and hosted by @monzo! I will share my learnings from fine-tuning @MistralAI 7B using synthetic data. Bring questions!
Sign up here https://t.co/rpIFjVm4h5
@e2b_dev's Agent Protocol looks really promising! Excited about its impact on multi-agent architectures like the Plan-Validate-Solve one of https://t.co/0C5cznVZHm 🤙
Btw, amazing to see Lemon Agent now being part of their Awesome AI Agents landscape 🥳
Spice up your AI game with Genoss GPT! 🌈🐂
Imagine swapping @OpenAI with any LLM, all in a single line.
- Switch to LLama V2 from @Meta? Done
- Experiment with GPT4ALL from @nomic_ai? Easy
- Custom model on @huggingface? Absolutely
Made w/ @langchain & @huggingface
🧵👇
Launching Lemon Agent! 🍋 https://t.co/0C5cznVZHm
🤝 Separation of tasks and human-in-the-loop interactions
🌌 Unlimited configuration options
👀 UI flexibility
🎨 [Soon] Model & framework agnostic operations
Huge thanks @ivanmartit, @pelaseyed and @_StanGirard for your input!