March 2026

The Hidden Cost in Every AI Agent You’ve Ever Built

The Hidden Cost in Every AI Agent You’ve Ever Built

⋅

And how GPT-5.4’s Tool Search is the first real fix for it. Here’s something I don’t see talked about enough in the AI/ML space. Every time you build an agent — doesn’t matter if it’s LangGraph, CrewAI, AutoGen, custom orchestration — you’re doing something quietly expensive in the background. You’re stuffing every tool definition into the context window upfront. Every function schema, every description, every parameter list. All of it. Before the model has even read your user’s message. Most of the time, the agent […]

Ver mais

Liked Liked
Data Collection Methods

Data Collection Methods

⋅

Choosing the Right Approach for Your Research Every data project starts with a fundamental question: where will the data come from? The answer shapes everything that follows. Choose the wrong collection method, and you might end up with biased samples, incomplete information, or data that simply does not answer your research question. Choose wisely, and you build a foundation that supports valid insights and actionable conclusions. Data collection is not a one-size-fits-all endeavor. A market researcher surveying customer satisfaction […]

Ver mais

Liked Liked

Towards Practical World Model-based Reinforcement Learning for Vision-Language-Action Models

⋅

Vision-Language-Action (VLA) models show strong generalization for robotic control, but finetuning them with reinforcement learning (RL) is constrained by the high cost and safety risks of real-world interaction. Training VLA models in interactive world models avoids these issues but introduces several challenges, including pixel-level world modeling, multi-view consistency, and compounding errors under sparse rewards. Building on recent advances across large multimodal models and model-based RL, we propose VLA-MBPO, a practical framework to tackle these problems in VLA finetuning. […]

Ver mais

Liked Liked

Bayesian Learning in Episodic Zero-Sum Games

⋅

We study Bayesian learning in episodic, finite-horizon zero-sum Markov games with unknown transition and reward models. We investigate a posterior algorithm in which each player maintains a Bayesian posterior over the game model, independently samples a game model at the beginning of each episode, and computes an equilibrium policy for the sampled model. We analyze two settings: (i) Both players use the posterior sampling algorithm, and (ii) Only one player uses posterior sampling while the opponent follows an […]

Ver mais

Liked Liked

Interpretable Operator Learning for Inverse Problems via Adaptive Spectral Filtering: Convergence and Discretization Invariance

⋅

Solving ill-posed inverse problems necessitates effective regularization strategies to stabilize the inversion process against measurement noise. While classical methods like Tikhonov regularization require heuristic parameter tuning, and standard deep learning approaches often lack interpretability and generalization across resolutions, we propose SC-Net (Spectral Correction Network), a novel operator learning framework. SC-Net operates in the spectral domain of the forward operator, learning a pointwise adaptive filter function that reweights spectral coefficients based on the signal-to-noise ratio. We provide a theoretical […]

Ver mais

Liked Liked

Terrified Comments on Corrigibility in Claude’s Constitution

⋅

(Previously: Prologue.) Corrigibility as a term of art in AI alignment was coined as a word to refer to a property of an AI being willing to let its preferences be modified by its creator. Corrigibility in this sense was believed to be a desirable but unnatural property that would require more theoretical progress to specify, let alone implement. Desirable, because if you don’t think you specified your AI’s preferences correctly the first time, you want to be […]

Ver mais

Liked Liked
Modern Questions vs Ancient Language

Modern Questions vs Ancient Language

⋅

Why My RAG System Struggled with the Bible — and How I Fixed It Introduction I was building a Retrieval-Augmented Generation (RAG) system for a Bible question-answering application when I ran into a surprising problem. The system worked well for direct verse lookups, but when users asked natural questions like “Why do people suffer according to the Bible?”, the retrieved passages were often only loosely related. The issue turned out to be neither the model nor the vector database — it was a more […]

Ver mais

Liked Liked
Turbo Pascal 3.02A, deconstructed

Turbo Pascal 3.02A, deconstructed

⋅

Turbo Pascal 3.02A, deconstructed In Things That Turbo Pascal is Smaller Than James Hague lists things (from 2011) that are larger in size than Borland’s 1985 Turbo Pascal 3.02 executable – a 39,731 byte file that somehow included a full text editor IDE and Pascal compiler. This inspired me to track down a copy of that executable (available as freeware since 2000) and see if Claude could interpret the binary and decompile it for me. It did a […]

Ver mais

Liked Liked
A Complete Guide on Data: How to Fine-Tune Correctly, Part 3

A Complete Guide on Data: How to Fine-Tune Correctly, Part 3

⋅

Chat templates, synthetic data pipelines, deduplication, and the exact failure modes that kill fine-tuned models before training even starts Generated using notebookLM The strangest fine-tuning failures are the ones where everything looks right. The loss curve descends cleanly. The evaluation metrics hold. The training run completes without a single NaN. You load the model, run an inference, and the output is wrong in a way that is hard to even categorize. Not confidently wrong, not randomly wrong, but structurally […]

Ver mais

Liked Liked
Jury finds Musk owes damages to Twitter investors for his tweets

Jury finds Musk owes damages to Twitter investors for his tweets

⋅

On Friday, a jury in California determined that Elon Musk had misled investors in Twitter via public statements that depressed the price of the company’s stock ahead of Musk’s purchase of the service. Because this was a class action lawsuit, Musk is likely to owe damages to a huge range of investors—payments that may ultimately reach billions of dollars. In the lead-up to Musk’s ultimate purchase of the social media platform, he made a number of comments on […]

Ver mais

Liked Liked

Load More

↑

Digitado © 2026