BusinessCareers

Smart Brevity

Jamiepluscoffee

My way of learning

Play

Open in

18 episodes
Avg 12 min
English

Episodes 18

Newest Oldest Longest

All18 Unplayed18 In progress0 Played0

Today · 25 min
AI Systems, Week 2: How much does the interface shape the model?
A five-week course for understanding what comes from the model, what comes from the harness, how agent systems are evaluated, and how to design architectures that remain adaptable as models change.
- Transcript
#11
Yesterday · 7 min
MailDump 11: Thirteen Out of One Hundred
Coding agent benchmarks versus reality, agents as the new reader, RSU tax rules, supplements versus reprogramming, hiring DNA, and an archive thread on conservation.
- Transcript
#5
Saturday · 8 min
Smart Brevity 5: Guarantees come from structure, not from asking nicely
Guarantees come from structure, not from asking nicely; Waypoint-gated distillation: transfer only the RL shifts that early and late training agree on; Decision-native models: null-aware, order-proof, scalable, and cacheable by construction; Thinking iner
- Transcript
Tuesday · 24 min
AI Systems, Week 1: Where does an agent's behaviour come from?
A five-week course for understanding what comes from the model, what comes from the harness, how agent systems are evaluated, and how to design architectures that remain adaptable as models change.
- Transcript
#10
October 11 · 7 min
MailDump 10: Practice Rooms for Robots, and a Cliff Already Born
World models get a price tag, agents get leashes, colleges meet demographic arithmetic, and the archive asks where the body fits in thinking.
- Transcript
#4
October 10 · 7 min
Smart Brevity 4: The signal is narrower than it looks: recency posing as memory, failures that teach, less supervision that wins
The signal is narrower than it looks: recency posing as memory, failures that teach, less supervision that wins; Memory-need stratification: pooled memory scores mostly measure recency; Memory framing: calling context "memories" changes how often it gets
- Transcript
#9
October 4 · 7 min
MailDump 9: Colleges Close, Solar Goes Red, and Who Checks the Work
Colleges closing on demographics, red-state solar, the 1% next door, offices and software decoupling from headcount, science after cheap thinking, who checks AI work, and an archive thread on founder speed.
- Transcript
#3
October 3 · 7 min
Smart Brevity 3: Your training signal has a shape, and it can lie: who teaches, how loudly, and who checks
Your training signal has a shape, and it can lie: who teaches, how loudly, and who checks; On-policy distillation scaling: small RL teachers can lift bigger students past themselves; Multi-teacher distillation: normalize each teacher's feedback spread or
- Transcript
#8
September 27 · 9 min
MailDump 8: What Endures When the Model Doesn't
Tech incumbents buy bargaining power instead of betting on one model, climate lawyers turn attribution science into a courtroom weapon, and business, money, and product writers convert fuzzy judgment into fixed rules. An archive dive asks whether AI self
- Transcript
#2
September 26 · 8 min
Smart Brevity 2: Measure the decision, not just the outcome
Measure the decision, not just the outcome; Regularized harness self-improvement: auto-tuned agent harnesses overfit unless edit size, specificity and bloat are constrained.; Decision-fork evals: grade an agent's choices at branch points; longer thinking
- Transcript
#7
September 21 · 8 min
MailDump 7: The Price of Proof, the Cost of Attention
A costly compute proof meets a cheap architecture; a forward deployed engineer's failure and a shadowing technique share a root cause; Oura's IPO and Big Tech's bundling disagree on wearable moats; and design mail shows craft judgment reasserting itself o
- Transcript
#6
September 6 · 11 min
MailDump 6: What Actually Landed
A catch up edition: AI writers agree the model itself is fading in importance, product writers converge on taste as the scarce skill, a UK campaigner and a US grid analyst reach the same fix for local power, a backdoor Roth quirk, and a five year archive
- Transcript
#1
September 5 · 10 min
Smart Brevity 1: Scaffolding is where this week's gains are, and it is dumber than assumed
This week in AI, taught properly: skill libraries, harness portability, KV cache eviction, cheaper agent evals, and where the scaffolding gains really come from.
- Transcript
#5
September 3 · 7 min
MailDump 5: Debt, Discipline, and Deciding Before the Storm Hits
A catch-up edition: is the AI money machine a bubble or the real thing, a five-step way to cap AI spending, climate finance pivoting toward land restoration, AI's indirect labor effects, and an archive dive into deciding before pressure hits, from Finding
- Transcript
#4
August 24 · 31 min
MailDump 4: The Podcast Roundup
The podcast roundup: two weeks of 20VC and More or Less, briefed.
- Transcript
#3
August 16 · 12 min
MailDump 3: The Half-Cent Model, the Return of the Room, and Carbon Removal's Reckoning
This week's dump sets a lab price war against one practitioner's real savings, checks Scott Galloway's four-part case for offline experience, watches two carbon removal newsletters split on who has to be the first buyer, and pulls a twelve-year idea-valid
- Transcript
#2
August 9 · 11 min
MailDump 2: Work What You Have, the People Bottleneck, and the Moat AI Can't Fake
This week's MailDump digs through a still-stalled newsletter backlog for what three growth operators, five climate voices, and two health researchers each independently discovered on their own terms.
- Transcript
#1
August 3 · 10 min
The Grid, the Bubble, and Two Business Truths
A debut catch-up on what your newsletters have been saying: the power grid as the real AI bottleneck, whether the boom is 1999 again, and two truths about how businesses actually grow.
- Transcript

Showing 1–18 of 18 episodes