Meridian Conversations

Ninety minutes with the person who ran the experiments.

Not the lab director, not the founder: the first author. One paper, the path that led to it, the experiment that nearly failed, the questions the reviewers asked and what comes next. Video and audio, with a full transcript linked to the paper.

Season oneSubscribe

Why authors

The most-listened technology shows interview the people who run companies and fund them. That is valuable, and it is not the same thing as hearing from the researcher who built the result. Meridian is for that conversation.

The work, not the reputationWe choose papers, then invite their authors. A first-year PhD student with an oral paper outranks a famous name with a minor one.
Technical, and still listenableQuestions are prepared with researchers who have read the paper. Explanations are for an informed listener, not a specialist only.
Checked against the paperEvery finding mentioned is linked to the table, figure or section that supports it.
Asia in the frameRecorded in Singapore, with a deliberate share of authors from labs across Asia, whose work is under-heard in English-language media.
The failures tooEvery episode asks what did not work. Negative results rarely make the paper; they are often the most useful part of the conversation.
Built to lastEpisodes are organised by venue, year and theme, so the archive becomes a reference to the field, not a feed.

Season one

Twelve themes from this year’s programme. Guests are announced on release.

01ICLR · NeurIPS

How models learn to reason

What actually changes inside a model trained to think step by step, and what the evaluations miss.

Guest announced on release
02SIGIR · ACL

Retrieval, memory and long context

When a model should look something up, when it should remember, and why long context did not end retrieval.

Guest announced on release
03ACL · EMNLP

Language models for Asian languages

Tokenisation, data and evaluation for Japanese, Chinese, Korean, Thai, Bahasa and the languages between them.

Guest announced on release
04ICML · COLM

Agents that use tools

Planning, tool use and recovery from error, and how to tell an agent benchmark from a demo.

Guest announced on release
05NeurIPS D&B · EMNLP

The science of evaluation

Contamination, judges that are models, and what a benchmark is for once models saturate it.

Guest announced on release
06ICML · MLSys

Training and serving efficiently

Quantisation, distillation, mixture-of-experts and the economics that decide which ideas ship.

Guest announced on release
07CVPR · ICCV/ECCV

Seeing, reading and listening

Vision-language models, documents as images, video understanding and speech.

Guest announced on release
08EMNLP · SIGIR

Documents, law and professional work

Contracts, filings and forms: the unglamorous documents where language models meet liability.

Guest announced on release
09NeurIPS · ICML

AI for science

Proteins, materials and weather: what transfers from language models to scientific discovery and what does not.

Guest announced on release
10ICLR · NeurIPS

Alignment and safety research

Interpretability, oversight and robustness, explained by the people running the experiments.

Guest announced on release
11COLT · ICML

Why deep learning works

Scaling, generalisation and optimisation: the theory catching up with practice.

Guest announced on release
12CoRL · CVPR

Robots and embodied learning

Learning from demonstration, simulation and the gap between a lab arm and a warehouse.

Guest announced on release

Every episode includes

  • Video (YouTube) and audio (all major podcast apps)
  • Chapters, so you can jump to the reviewers’ questions or the limits
  • A full transcript with timestamps and links to the paper, code and data
  • A one-paragraph summary in plain language, and the sixty-second take
  • Disclosures: travel costs covered, any sponsorship, any relationship to Magellan Global companies

The episode format Editorial standards