AI Radar

Building prompts with LLMs

I’ve been into AI image and video generation for a while, and I’ve been a bit obsessed with H3 for the past two months. With the right references and prompt, it feels like I can make almost anything. Writing that prompt is the tedious part, so I usually hand it to an LLM. For most of those two months, I used closed models, especially Grok (for obvious reasons). My problem is that they rarely add much on their own (no creativity). If I don’t explicitly ask for an emotion, expression, or action, they tend to leave it out. I've had generations where a character is stiff as a tree for 15 seconds…

How NVIDIA GPUs Help Accelerate OpenAI’s GPT-6 Astra Ultrafast

GPT-6 Astra Ultrafast, running on NVIDIA Blackwell GPUs, is available now in the OpenAI API and to eligible ChatGPT Work and Codex users. Accelerated by inference optimizations through OpenAI’s models that tap into the capabilities of the NVIDIA Blackwell architecture, Ultrafast offers up to 8x faster token generation than the Astra Standard mode. For developers, […]

Foundations of Large Language Models

This is a book about large language models. As indicated by the title, it primarily focuses on foundational concepts rather than comprehensive coverage of all cutting-edge technologies. The book is structured into six main chapters, each exploring a key area: pre-training, generative models, prompting, alignment, inference, and reasoning. It is intended for college students, professionals, and practitioners in natural language processing and related fields, and can serve as a reference for anyone interested in large language models.

A city-building game in which the city would prefer you didn't

After losing my job and 5 years of doing all the marketing tricks I read everywhere, I got massive hype just by sending emails, and nothing else worked (city builder game at 15k wishlists)

The wishlist count went from the valley of despair, with close to 0 wishlists every day, to 400 on the day of the demo launch, and then peaked at 1,400 wishlists in a day. And now I'm getting 100+ wishlists per day without doing anything special. See here . TL;DR: The emails sent to YouTubers for my 3rd and last game on Steam, DreadHaven: The Last Colony , worked, and it skyrocketed everything. Not my Shorts, not my social media posts, just the emails. Some people will say, "Yeah, just make a good game, duh", but I still feel like the 5 years of failed attempts helped me achieve my overnight…

The issues with being able to reverse engineer games in priority order.

Professional gamedev here (17yrs), and I just wanted to give you my perspective on the major issues the industry faces from this emerging problem, from what I think is the most problematic to the least. People reverse-engineering games and then re-releasing the games for free in a browser or on some other platform. Imo this is simply piracy. There's some precedent for this from the music industry, which has had to tackle similar problems for years. I can't imagine this lasting too long; I think the IP holders will bring the legal hammer down hard on people doing it. Then there's the argument…

Build your own decision model

Asana cuts model costs 76x in browser tests with GPT-6.1 Sol

Using GPT-6 Astra in Codex, Asana made its browser agent 76x cheaper and 5x faster in tests to offer customers more capable models.

REMORY: Learning Residual Memory for Context Compaction

Long-horizon agents compact their history to continue within a finite context window, but a textual summary alone may not support every subsequent decision. We introduce REMORY, a neural memory network that supplements the summary with a bounded sequence of soft memory tokens. Given the history and summary, the network learns to generate tokens that help a frozen LLM approximate the continuation it would produce with the full history. The tokens are conditioned on the summary and appended after it, forming an analogue of a residual connection along the sequence dimension. On SummHay, REMORY…

The Agent Said It Was Done. The Database Disagreed.