The critic: how the briefing is rated, every week

Every Saturday morning the briefing is scored against the latest episode or issue of ten of the most popular tech and AI shows, with one fixed rubric. The aim is practical: find the few changes that would most improve the briefing for its readers and listeners, decide each week which to make, and see whether they worked.

The rubric

Ten characteristics, each scored 1-10 (1-3 poor, 4-6 adequate, 7-8 strong, 9-10 best in class). The overall score is the weighted sum divided by 10, out of 100. The weights are for a daily AI briefing for investors, set from what audiences consistently praise and criticise in reviews, rankings and press coverage of these shows: signal-to-noise first, then accuracy and depth; brevity and a fixed structure for a daily; conviction that is fair to the other side; voice as the moat; polish counts but comes second.

Characteristic	Weight	What is measured
Content selection	18	Signal-to-noise: the right stories, no filler
Accuracy & sourcing	15	Traceable claims, links to primary sources, no contradictions within an edition
Depth & insight	14	The "so what" for capital and markets
Timeliness & cadence	10	Fresh and dependable enough to become a habit
Structure & pacing	10	A predictable order, and a length that fits the format
Opinion / point of view	9	Conviction that is fair to the other side
Distinctiveness & fit	7	Hard to get elsewhere; built for this audience
Writing / script	6	Clarity; written for the ear in the podcast
Tone & voice	5	Personality, chemistry, register
Production	6	Audio fidelity, music, cold open, interludes, transitions, editing; design for a newsletter

Production problems a listener notices cap the score, whatever the other characteristics.

The benchmark shows

Chosen in October 2026 for popularity (chart positions, subscriber and audience figures, rankings) and for relevance to a daily AI briefing for investors: Dwarkesh Podcast, Stratechery, The AI Daily Brief, Machine Gods (NPR), Latent Space, The Neuron, Acquired, All-In, TLDR AI and the Lex Fridman Podcast. Two changes from the first list: Machine Gods replaced Hard Fork when its hosts moved there, and TLDR AI (about 1.15M subscribers) replaced The Batch, which publishes no audience figures; The Batch was kept as a bonus benchmark in the first run.

Long interview and history shows (Acquired, Lex Fridman, All-In) score low on fit and length for an investor daily. That reflects the use case, not their quality.

How a weekly run works

1. The ten shows. Each show's latest episode or issue is scored. Nobody listens to the audio: scores come from transcripts, show notes, chapter lists, reviews, press coverage and production credits, and each score carries a confidence note. Newsletters are scored on the issue itself.
2. The briefing. An independent critic reviews the past seven editions: the written edition, the podcast script, the deep dive, the run statistics, takes and corrections. The written edition and the podcast are scored separately where they differ (structure, writing, tone, production), then combined. Measurements are included: word counts, durations, sections and their order, how often a story repeats across days.
3. The feedback. Five strengths to keep, and ten suggestions ranked by expected effect on the score. Each suggestion carries evidence (a quote or a measurement) and two to five concrete options, with one recommended, plus a first batch with the most impact for the effort. Free, local fixes come first.
4. The decision. The owner receives a private email (below) and chooses what to build that week. Changes go through the normal release process, and the next week's run checks whether they moved the scores.

Outputs

• The weekly email (owner only, pipeline/critic.py): our score and rank; the top five suggestions with the recommended option; a bar chart of our score against the field average and the best show for each characteristic, marking where we are below the field; the overall ranking; our trend week by week; what shipped that week; strengths to keep. The charts are plain HTML, so they render in any mail client.
• The record: data/critic/<date>.json (scores, suggestions, options) and data/critic/<date>.md (the full narrative report), kept in the private archive. The trend is read from these files.
• Commands: ./np critic preview | email | list, and the run-queue job critic-email.

Limits

• Scores are informed judgements by a model, not measurements. Different reviewers score different shows, so differences of about 3 points are noise; the trend over several weeks matters more than one week.
• Audio can't be heard, so production scores rest on documented evidence and carry lower confidence.
• Only the latest episode or issue of each show is scored, so an unusual week moves its score.
• The rubric is weighted for this briefing's use. It is not a general ranking of the shows.

Latest results

Generated at build time from the critic run of 2026-10-31.

Rank	Show	Score	Rated
1	Stratechery	77.0	The End of the Internet (2026-10-26)
2	The AI Daily Brief (newsletter)	74.4	How the Most AI-Pilled Company in the World Works (2026-10-30)
3	The AI Daily Brief	73.0	How the Most AI-Pilled Company in the World Works (2026-10-30)
4	Latent Space	71.6	Why Agents Love Supabase (2026-10-28)
5	All-In	70.8	Anthropic S-1 Shocker, Greatest Business Ever? Google's $500 (2026-10-31)
6	Acquired	70.6	Seinfeld (Fall 2026, Ep 3) (2026-10-18)
7	AI Briefing (us)	69.6	the week's editions (2026-10-31)
8	Machine Gods (NPR)	69.4	OpenAI's Latest Discoveries Spark a Backlash + Help! My Kid (2026-10-30)
9	The Neuron	64.5	Mistral released a 1T-parameter open model (2026-10-30)
10	Lex Fridman Podcast	62.5	#503 AI Crisis in Math (2026-10-26)
11	Dwarkesh Podcast	61.2	Robert Caro, the master biographer, reflects at 91 (not AI) (2026-10-30)
12	TLDR AI	57.8	Daily issue (OpenAI Model Spec; Mistral open weights) (2026-10-30)

Characteristic (weight)	Us	Field average	Best
Content selection (18)	6	7.0	8 (Stratechery)
Accuracy & sourcing (15)	7	7.2	9 (Acquired)
Depth & insight (14)	8	6.0	8 (All-In)
Timeliness & cadence (10)	8	7.3	9 (The Neuron)
Structure & pacing (10)	7	7.3	8 (The AI Daily Brief (newsletter))
Opinion / point of view (9)	7	6.3	9 (Stratechery)
Distinctiveness & fit (7)	8	5.8	8 (Stratechery)
Writing / script (6)	6	7.3	9 (Stratechery)
Tone & voice (5)	6	7.5	8 (The AI Daily Brief (newsletter))
Production (6)	6	7.2	9 (Acquired)

Our overall score by week: 2026-10-24 62.2 · 2026-10-31 69.6