Max

Full-Stack Engineer, CTO & Co-Founder

[Max @max_gosk]

We will be developing more and more software for agents to use, most companies will be building for agent companies rather than direct to user. Onboarding will be seamless

[Tulip King 🌷 @tulipking]

i look forward to our chinese brothers liberating the knowledge from within fable-5 and selling it to me at 5% the cost & 2x the speed

[Max @max_gosk]

@trq212 I’ve used html outputs for almost a year and then reverted to MD. Models (gemini’s at least) tend to default MD for tables or mixed outputs(even when agressively prompt to output html); breaking rendering like <br> inside a markdown cell etc. Now using tools for inline html instd

[Max @max_gosk]

@JoannaStern Install gmail as an app from Chrome, settings > install as app or sth. Works for any website.

[Naval @naval]

A lot of software is about to get a lot better, right before it becomes unnecessary.

[Melvin Vivas @donvito]

Wake me up if GPT 5.2 has beaten Opus 4.5

[Max @max_gosk]

@levelsio Going there for new years, share more interesting findings!

[Max @max_gosk]

@OfficialLoganK Miss the verbosity from 2.5

[Google AI Studio @GoogleAIStudio]

gemini 3 pro

• our most intelligent model yet
• SOTA reasoning
• 1501 Elo on LMArena
• next-level vibe coding capabilities
• complex multimodal understanding

available now in Google AI Studio and the Gemini API

[Max @max_gosk]

@OfficialLoganK @grok how this applies to the vercel AI sdk

[Max @max_gosk]

Chinese mobile apps are a different breed

[Max @max_gosk]

Thanks @posthog for the startup kit!! + the shirt, ready for Saturday coding 🦔🦔

[Max @max_gosk]

@DanielLockyer Got DMd by multiple people asking how I got such high scores 🤣🤣

[Max @max_gosk]

@rqobela Php

[Max @max_gosk]

@DanielLockyer Thought about this, but how you can detect picture of a picture?

[Max @max_gosk]

@Mrbankstips Buy tesla

[Max @max_gosk]

@levelsio A good wrap has avocado

[Max @max_gosk]

@pmitu Team 100%

[Max @max_gosk]

We have been doing this for a while at Lawdify, we call it “Case Brain”, recommended 👍🏻

[Rohan Paul @rohanpaul_ai]

This paper introduces a new method called Agentic Context Engineering (ACE).

It helps language models improve by updating what they read and remember, instead of changing their core weights.

Normal methods that edit prompts tend to make them too short and lose important details over time.

ACE fixes this by treating the model’s context like a growing notebook that keeps and organizes useful strategies.

It has three parts: a Generator that works on tasks, a Reflector that learns from mistakes, and a Curator that updates the notebook with helpful lessons.

Instead of rewriting everything each time, ACE makes small, local edits, which keeps learning fast and preserves past knowledge.

It also trims repeated or useless points so the context stays clean and focused.

In tests with coding agents and financial reasoning, ACE gave about 10% higher accuracy and cut learning time by about 87%.

It even worked well without labeled training data, using only real-time feedback from how tasks ran.

----

Paper – arxiv. org/abs/2510.04618

Paper Title: "Agentic Context Engineering: Evolving Contexts for Self-Improving Language Models"

[Max @max_gosk]

@gunsnrosesgirl3 Shouldnt the door open inwards instead?

[Max @max_gosk]

I love typescript and LLMs