GPT 5.4 will be another GPT 5.3 moment.
TOOLS
-
SAGE-RL: Simple RL Modification for Better Training
By
–
then they fold it into training with SAGE-RL. dead simple modification: in standard reinforcement learning (GRPO), you sample 8 responses per question. SAGE-RL replaces 2 of those 8 with SAGE-generated samples. the other 6 stay normal. one-line code change. the model learns to
-
GPT-5.3-Codex-Spark Available on Poe Platform
By
–
Try it alongside every top model in the Poe app and API today at https://
poe.com/GPT-5.3-Codex-
Spark
…. Pricing is 10 points/message to ensure broad access. -
GPT-5.3-Codex-Spark now available for fast code generation
By
–
GPT-5.3-Codex-Spark is now available on Poe in a limited research preview! Poe is one of the first places where this model is available today. Enabled by Cerebras’s high-performance chips, OpenAI's newest lightweight coding model is optimized for ultra-fast code generation
-
Relying on AI for Legal Advice: A Positive Experience
By
–
Holy sh**! I’ve been overwhelmingly relying on AI for legal advice over the last few months. I had tried using a lawyer before, but he cost me $$$ and gave me crappy advice.
-
Peter Steinberger conversation now available in German with AI dubbing
By
–
My conversation with Peter Steinberger (@steipete) is now translated & dubbed into German.
— Lex Fridman (@lexfridman) 4 mars 2026
Huge thank you to ElevenLabs (@elevenlabsio) and @matiii for making it happen.
It's available here and on YouTube where you can switch audio tracks by clicking the gear icon > Audio… pic.twitter.com/KqLb53KFBzMy conversation with Peter Steinberger (@steipete) is now translated & dubbed into German. Huge thank you to ElevenLabs (@elevenlabs) and @matiii for making it happen. It's available here and on YouTube where you can switch audio tracks by clicking the gear icon > Audio Tracks > select German. I have a lot of hope for this application of AI to break down barriers that language creates, and no one does it better than @elevenlabs. The whole ElevenLabs team has been really fun & inspiring to work with.
→ View original post on X — @lexfridman, 2026-03-04 00:08 UTC
-
Google Launches Retail Business Agent with Flash-Lite Model
By
–
Second, using 3.1 Flash-Lite in @GoogleAIStudio, we built a retail business agent that plans and executes multi-step tasks for you like reporting and dashboard automation – saving you time to focus on the things that matter most.
— Google AI (@GoogleAI) 3 mars 2026
Beginning today, the model is rolling out in… pic.twitter.com/XOyQ6nGZROSecond, using 3.1 Flash-Lite in @GoogleAIStudio
, we built a retail business agent that plans and executes multi-step tasks for you like reporting and dashboard automation – saving you time to focus on the things that matter most. Beginning today, the model is rolling out in -
AI builds websites instantly, empowering everyday users with agentic apps.
By
–
True, but wait until they can tell an AI to do something, like build them a new website, and it does in front of them. That's why I love @wabi so much. It's approachable for everyday people, but it builds highly useful agentic applications automatically.
-
LLMs Should Generate 3D Code Like They Generate Executable Code
By
–
We don't expect LLMs to multiply numbers or sort lists directly within their output token stream. Instead, we ask them emit code and execute it in a separate runtime. Why predict the opposite outcome for simulating interactive worlds? worldlabs.ai/blog/3d-as-code
-
MLOps Stack and AI Agents: Critical System Design Considerations
By
–
Very much succinct write up on each tech stack! 1) MLOps: very much important to note done tools you are gonna use and their tradeoffs. system design is critical to know where you are gonna fit AI or not. everything is AI first now a days. 2) Agents: Most of the critical tool