Towards Data Science

virus spreading through big tech. Engineers are being judged, directly or indirectly, by how much AI they can consume. More tokens, more output, more compute. Some companies even had leaderboards. It’s the 2026 version of ranking engineers by lines of code. Tokenminning is the antithesis of tokenmaxxing. Token efficiency becomes increasingly important as your usage grows. Every unnecessary token increases cost, latency and complexity.

“We don’t write prompts anymore. We design loops.” — someone at Anthropic in June 2026 agent loop, self-critique did no better than doing nothing. A deterministic, source-anchored verifier cut the hallucination rate roughly in half. The line is from a few weeks ago and already feels true. We stopped tuning one perfect prompt and started building systems that try, check their own work, and improve over several steps. A model that can revise is worth more than a model that answers once and stops.

way to understand the new time-series foundation models, so I picked a recent one I could run. t0-alpha is a 102M-parameter probabilistic forecaster from The Forecasting Company, released in June 2026. The Forecasting Company published the weights openly under Apache-2.0, which is what makes this reproduction possible: the model is small enough to run on accessible hardware, and it ships with GIFT-Eval results that can be checked outside the original lab.

companion to Enterprise Document Intelligence, the series whose philosophy is laid out in Amplify the Expert. It zooms in on brick 2 (question parsing) of the four-brick architecture and surfaces the lessons most tutorials skip. Most RAG tutorials skip question parsing. The user’s string goes straight to retrieval, cosine runs on top-k, and the model gets handed whatever came back. We do not do that, for one reason: a user question is not a search query.

[1] examined how powerful machine learning can look deceptively convincing when the evaluation setup is flawed. However, in spatial prediction problems, such as real estate applications involving capital gains estimation, rent forecasting, or price prediction, the problem does not end with fixing temporal leakage. Even when time is handled correctly, models can still appear far better than they really are if spatial dependence, repeated-asset structures, and uneven regional coverage are ignored.

A humorous-but-real tour of ILCP-for-agents — a β-VAE compressor, an Xn-style transport, a gated MLP projector, and the unreasonably convenient realization that I had already solved this exact problem for 6G handovers. The agent-side V1 is the wiring; the receipts in this post are the 6G paper’s receipts, properly labelled, because honest writing is the whole point of this series. — the capstone — of the “Production-Grade Agentic Inference” series.

, memory has undoubtedly become critical resource. As the demand for memory and storage infrastructure driven by the AI boom has reached historical highs, companies like Micron Technology and Sandisk have attracted unprecedented attention and increased product prices by wielding strong pricing power.

to perform useful work can be relatively straightforward. For a simple application, a few lines of Python using the boto3 library and the Bedrock API may be all that’s needed. You set up access to an LLM, send a prompt and some input, receive a response, and return it to the user. That approach becomes harder to manage if your application has to take on more responsibilities.

companion to Enterprise Document Intelligence, a series whose stance is that enterprise RAG amplifies the expert, it does not replace them. The architecture follows from that: four bricks (document parsing, question parsing, retrieval, generation), each emitting typed pieces that converge on one LLM call. The industry now calls that practice context engineering. Scope here is the single-document case; corpus, conversation, and tool-call extensions are follow-up work.

in on interviews at my previous company, I used to think the behavioral part of the interview was the easy part. That if someone was technically proficient and could wow the interviewers with their analytical mind, they’d be the top contender for any job. Then I witnessed a highly technically skilled applicant lose out to a more socially competent one. Not because he lacked experience. He had done good work, and he obviously knew what he was doing.