Towards Data Science

The world's leading publication for data science and artificial intelligence professionals.

Submit an Article ✍️

[Towards Data Science @TDataScience]

Do you need your code to make quick, reliable decisions about images, not just describe them?

@taupirho shows how with hands-on Python examples using Jev’s new image support, from sorting product photos to spotting damage on a car. https://t.co/YbEChne0AO

[Towards Data Science @TDataScience]

AI agents can write your data loaders and run your experiments in a fraction of the time they required before. But as Fiona Victoria explains, “a missing observation is not the same as an absence,” and no amount of speed fixes data with a backstory. https://t.co/T3Oz6gXZhZ

[Towards Data Science @TDataScience]

What happens when your local LLM’s answer feeds straight into your pipeline and nobody checks it?

Shuai Guo shows how to add a second, fully local reviewer using a decision model, so your data never leaves your machine. https://t.co/JXBYrzIhmq

[Thomas Reid @taupirho]

JEV can now process images. Check out my latest @TDataScience article to find out how.

https://t.co/t22wJ8kTKx

[Towards Data Science @TDataScience]

Is a model that predicts 50/50 just indecisive, or might it be “the model confidently reporting that the answer is ambiguous"?

Maja Pavlovic introduces a relabeled MNIST dataset that captures how unsure humans are about each digit. https://t.co/yrbTL4bR8E

[Towards Data Science @TDataScience]

Daniel Nweke built a study planner where GPT-4o picks lessons from the YouTube API:

https://t.co/1V1jvCjTmA

[Towards Data Science @TDataScience]

When neurons seem to “avoid” microglia in a glial scar, the truth may be simpler: “nothing repels anything; there is simply nobody left to be close to.”

In Part 3 of his spatial analysis series, Daniel Manrique Castano shows how to tell tissue loss from real interaction. https://t.co/r2y1bpvfaq

[Towards Data Science @TDataScience]

Tell a small language model “That’s wrong” again and again, and one started rotating through the options — A, B, C, D, A, B, C, D — “as if ‘wrong’ meant ‘next.’”

@mangal_utkarsh pushed back 22,500 times on three open models to see what happens to a correct answer. https://t.co/gfpZ0LLJ4D

[Towards Data Science @TDataScience]

Assistant, copilot, agent: vendors use them interchangeably, and Gartner expects over 40% of agentic AI projects to be cancelled by 2027. Gursimar Singh uses the example of a palace-hotel wedding week to show what each one actually does, and when you need orchestration. https://t.co/8KHJFg8Buu

[Towards Data Science @TDataScience]

macOS 27 comes with a free, private LLM in the Terminal. But as Irfan Alghani Khalid found out, valid JSON “only guarantees the shape of the output, not its truth.” https://t.co/USuNBxYhZT

[Towards Data Science @TDataScience]

Can a model pick the right tissue type from a description it was never trained on?

Lambert Leong builds a Jev-style decision model for pathology images, compares it with Qwen2.5-VL and trained classifiers, and shows where that flexibility helps and where it doesn’t. https://t.co/QF9ElVUCAw

[Towards Data Science @TDataScience]

Should that rule live in a Skill or an MCP server? Gal Arav argues the real question is what stops the LLM from ignoring it.

https://t.co/MERRDYdPj1

[Towards Data Science @TDataScience]

Google's Gemini 4 Argon brings a million token output window, but its benchmarks don't win everything. Benjamin Nweke looks at where it shines, where it trails, and who it actually suits.

https://t.co/xMqF1GObzL

[Ibrahim Salami (Ibby) @ibbysalam]

I recently build a job market pipeline that extracts data from two sources (Hackernews and RemoteOk).

So I decided to enrich it with AI. Now I can easily know: the skills the jobs request for, level of seniority and if it’s open to Nigeria)

It’s published on @TDataScience

[Gal Arav @GA4198498563411]

How effective are Python test suites really?

I planted 2,312 bugs 🐛 in 50 libraries and measured what survived. The youngest projects missed 41% of injected faults. 🚀✨

https://t.co/YR5mAO9yCQ

Thanks @TDataScience team for publishing it, made it as an Editor's Pick!

[Towards Data Science @TDataScience]

Routing tasks to a cheaper model sounds like an easy way to save money, but switching mid session can wipe out those savings. Nhu Hoang tests why, and shows which routing approach actually keeps costs down.

https://t.co/9hnP91ETBI

[Towards Data Science @TDataScience]

Not sure what your coding agents are shipping?

@EivindKjos shows how we can use Claude Code to write feature memos, review pull requests, and build visual HTML explainers. https://t.co/tpl6Xeau8x

[Towards Data Science @TDataScience]

If you're working with AI agents and keep running into issues around which component in your system should make decisions, don't miss Gursimar Singh's primer on harness engineering. https://t.co/mBPKLseqEi

[Towards Data Science @TDataScience]

If you're interested in basketball, sports analytics, or even just solid machine learning project walkthroughs, Pol Marin's recent article is a must-read. https://t.co/r7jldHHqlJ

[Towards Data Science @TDataScience]

How will humans redefine themselves in relation to work once AI takes over entire professions? And what should society at large do to make sure this transition works for people?

James O'Brien charts a pragmatic, levelheaded path towards a potentially sustainable future.
https://t.co/fMS1wxIl3Q