Gabriele Lanaro

The first public foundation models from @poolsideai just dropped on OpenRouter! Laguna M.1 and Laguna XS.2. Built from scratch for agentic coding and long-horizon work. Free for a limited time ⬇️

Today we’re releasing Laguna XS.2, Poolside’s first open-weight model. It’s a 33B total / 3B active MoE model built for agentic coding and long-horizon tasks. Trained fully in-house on our own stack. Runs on a single GPU. Released under Apache 2.0. Links 👇 Weights: huggingface.co/poolside/Lagun… API: platform.poolside.ai Blog: poolside.ai/blog/laguna-a-…

@LumaLabsAI I ran out of free credits, I can get the 30$ subscription how many credits does image generation with UNI-1 consumes? Is it available in the paid plan? Very difficult to understand from the website

Uni-1 is here! A new kind of model that thinks and generates pixels simultaneously. Less artificial. More intelligent.

@thsottiaux clear session or trigger compaction

With GPT-5.4 out. What should Codex ship or improve next?

@Kimi_Moonshot No thank you! Amazing work, fast and good at coding, writing and frontend

Kimi is currently the #1 on LLM Leaderboard (via OpenRouter) This marks our first time topping the all-time usage charts, a huge milestone for the team. A massive thank you to every developer and user who made this possible. 🫡

From two levels to a real B+tree (Part 2) open.substack.com/pub/teamkitche…

A bit later this week, but this is finally the moment everyone was waiting for. The new post about the WrongoDB journey, the first single-level, cute, little B+tree. open.substack.com/pub/teamkitche…

@thekitze Really good concept!

few weeks ago someone tweeted "vibe coding is like factorio for devs" so i decided to make a game interface for controlling agents

@svpino Yes codex is great and gpt-5.2-codex is a fantastic model (superior to opus in many circumstances)

Is anyone still bothering with Codex or Gemini CLI? Have we all already decided Claude Code is the winner? Nothing wrong with fooling around with other tools, but I'm a huge proponent of sticking with a single tool through thick and thin and trying to know it well. You just can't build mastery if you are switching tools every week. Anyway, Claude Code is the one for me.

@dabit3 Because human written code is of such high quality

Unpopular opinion - AI-generated code will be of higher quality than 99.9% of human-created code in a very short time frame (< 6 months)

@Prathkum Code organisation/design matters a lot, it makes it so the agent finds effective patterns already and is able to evolve the code base without ending with a ton of defects

My take on the quality of AI-generated code. Code is now cheap, fast, and generated at a rate that completely out of our ability to read, understand, or maintain it line by line. That alone breaks a lot of our old intuitions about code quality. Pre-AI, code quality mattered because humans were the bottleneck. We had to read, refactor, and maintain. Carefully reason about every line because writing code was expensive. If the end product behaves correctly, internal codebase matters far less than we have been taught to believe. This doesn’t mean quality disappeared. It just moved from code to:

– the spec
– the agent config
– the system prompt
– the feedback loops

@thdxr The difference matters less and less over time

ok i'll give you the answer to this - it's so simple and some people will find the truth irritating your first experience with one of these tools is magic your first experience with the other you might think it's broken ultimately no other detail matters

Where the Records Go: Slotted Pages open.substack.com/pub/teamkitche…

It's Monday! Another WrongoDB Post. How do you store variable sized record in databases? Where the Records Go: Slotted Pages open.substack.com/pub/teamkitche…

Why isn’t anybody undressing in better planning tooling best we have is some markdowns in the code base. But since coding is offload of an agent planning is actually what I spend most of my time doing

Alright I said it, gpt-5.2-codex max xhigh is best so far, yes better than opus. Prudent, precise, pushes back sometimes

When does the OS writes stuff to disk: my exploration with fs_usage open.substack.com/pub/teamkitche…

GLM-4.7 is here! GLM-4.7 surpasses GLM-4.6 with substantial improvements in coding, complex reasoning, and tool usage, setting new open-source SOTA standards. It also boosts performance in chat, creative writing, and role-play scenarios. Default Model for Coding Plan: z.ai/subscribe Try it now: chat.z.ai Weights: huggingface.co/zai-org/GLM-4.7 Tech Blog: z.ai/blog/glm-4.7

Make it wrong before you make it right, I'm vibing the wrong database in rust WrongoDB: a tiny wrong database (on purpose) open.substack.com/pub/teamkitche…

GPT 5.2 is now available in Cursor. Priced at $1.75/M input, $14/M output tokens.