JSON Halliday

CEO, Researcher and Speaker✨ dabbler 🚀

[OpenAI @OpenAI]

Building apps has never been easier.

With Sites, Codex can turn your work, ideas, and plans into an interactive website or app your team can explore, use, and share with a URL.

Rolling out to Business and Enterprise plans, before expanding more broadly.

[JSON Halliday @Jsonhalliday]

WorldSeed, Don't build a workflow. Seed a world.

rules + different agents + consequences -> emergence

Define roles, rules, private information, actions, and consequences. Then agents interact until useful artifacts emerge.

https://t.co/H0QZ4blGV9

[Dan Kornas @DanKornas]

Most AI/ML books are only useful if they change how you build.

Grokking Software Architecture is useful because it maps the topic to engineering work you actually have to operate.

The book covers:

• Deploy and scale beyond a demo
• Work from practical Python examples
• Understand the core system design choices
• Map the topic to real engineering workflows
• Avoid common production failure modes
• Build a stronger implementation plan

The production angle is the part I would pay attention to.

The demo is the easy part. The useful engineering work is making the system reliable once it touches real workflows.

Good fit for engineers building real AI systems and wanting a stronger mental model than another clean demo.

Link in the first comment.

[Dhanian 🗯️ @e_opore]

Backend Engineering Interview Questions:

1. What is backend engineering, and what are the responsibilities of a backend engineer?

2. Explain the difference between monolithic, microservices, and serverless architectures.

3. How do you design scalable backend systems?

4. What are the key principles of distributed systems?

5. What is horizontal scaling vs vertical scaling?

6. What is load balancing, and how does it improve system reliability?

7. What is the CAP theorem, and why is it important?

8. Explain strong consistency vs eventual consistency.

9. What is database sharding, and when would you use it?

10. What is database replication, and how does it work?

11. What is caching, and what caching strategies have you used?

12. What is the difference between Redis and Memcached?

13. What is a message queue, and why is it used?

14. Explain the difference between Apache Kafka and RabbitMQ.

15. What is event-driven architecture?

16. What are REST APIs, and how do they differ from GraphQL?

17. What is API versioning, and how do you implement it?

18. What is rate limiting, and why is it important?

19. What is authentication vs authorization?

20. Explain JWT-based authentication.

21. How do you secure backend APIs?

22. What are common backend security vulnerabilities, and how do you prevent them?

23. What is a database index, and how does it improve performance?

24. How do you optimize slow database queries?

25. What monitoring and logging tools have you used?

26. What is observability, and why is it important?

27. How do you troubleshoot production incidents?

28. What is fault tolerance, and how do you design for failures?

29. How would you design a URL shortener system?

30. How would you design a large-scale chat application?

Grab the Backend Engineering Ebook: https://t.co/t9mqUuRbjx

[Dan Kornas @DanKornas]

Training an LLM from scratch is easier to study when the whole path is in one repo.

Train LLM From Scratch is a PyTorch repository for learning how a transformer language model is built, trained, saved, and used for text generation.

It helps you move from “I understand attention on paper” to a runnable training pipeline by pairing model code with data download, preprocessing, config, training, and generation scripts.

Key features:

• Transformer components from scratch – separate PyTorch modules for MLP, attention, transformer blocks, and the final model
• Pile-based data path – scripts download The Pile files and preprocess JSONL.ZST text into tokenized HDF5 datasets
• Configurable training setup – model size, context length, heads, blocks, batch size, learning rate, and file paths live in https://t.co/zuPqaR3MhP
• Hardware guidance – README compares common GPUs for 13M and 2B-class training runs
• Generation workflow included – generate_text.py loads trained checkpoints and produces sample text outputs

It’s open-source (MIT license).

Link in the reply 👇

[JSON Halliday @Jsonhalliday]

GraphRAG still has a reasoning gap. This repo shows one way to train through it. Graph-R1 is the official implementation and resource repo for “Graph-R1: Towards Agentic GraphRAG Framework via End-to-end Reinforcement Learning.

https://t.co/xQqSTEB4d7

[Kyle Anthony Miller @kyleanthony]

I get to wake up and create stuff like this every day

[Charlie Holtz @charlieholtz]

Conductor + Browser + Agentation

[JSON Halliday @Jsonhalliday]

Massive Bug - Don't ever gamble in the new Expansion City Temis!

[DatModz @DatModz]

There is an Obol bug in @Diablo 4! Please fix it.

If you're not Blizzard and you want Mythic or Unique items, make sure to gamble your Obols at any Vendor outside of Temis!

[Diablo @Diablo]

The climb starts now 🔥

Ladder Season 14 is officially live in Diablo II: Resurrected - Reign of the Warlock.

[JSON Halliday @Jsonhalliday]

Instructions in markdown, output in html.