Build AI that actually works: https://heyclarity.dev/
Self Aligned Newsletter: https://www.selfaligned.me/

I sat down with my co-founder Jonathan to report from the front lines. This week China took the lead: Moonshot’s Kimi K3 became the first open-weight model to top the front-end code arena, beat the frontier labs at a fraction of the cost, and started a real fight in the White House over what open weights mean for American AI.

We get into Jensen Huang’s first-ever tweet and the open-weights letter he rallied half the Fortune 500 behind, why he treats model weights more like electricity than nuclear weapons, and the harder question underneath it: if distillation commoditizes frontier intelligence, do OpenAI and Anthropic still have a reason to build GPT-7?

We close in the meta: Cursor’s agent-swarm economics (the same task, six ways, from $1,339 to $20,057), Anthropic’s new rules of context engineering, a live demo of our Clarity harness, and the one command every builder should run this week.

If you’re a founder, engineering leader, or builder trying to make sense of the open-weight shift and what it means for your token budget and your moat, this one’s for you.

Timestamps:
00:00 Kimi K3 vs Claude on cost
00:19 The agenda
01:30 China’s open-weight bombshell
03:15 The industry reacts
07:15 Token budgets & the AI gateway
08:40 A Sputnik moment?
10:30 The Hugging Face attack
11:00 Language as a public utility
13:40 Jensen’s first tweet
14:00 Nvidia’s open-weights letter
20:40 Elon backs Jensen
22:00 American thought leadership
28:40 Can OpenAI & Anthropic win?
35:00 Plan frontier, build cheap
37:40 AI memes
40:30 Cursor’s swarm economics
46:00 The anti-slop taste skill
48:20 New rules of context engineering
54:00 Demo: the Clarity harness
1:01:00 Claude writes its own harness
1:04:00 The 6 workflow patterns
1:09:00 Claude Doctor
1:15:00 Wrap up

Follow Robert:
Instagram ► https://www.instagram.com/therobertta
TikTok ► https://www.tiktok.com/@therobertta_
LinkedIn ► https://www.linkedin.com/in/therobertta
X ► https://x.com/therobertta_
Substack ► https://robertta.substack.com/

Follow Jonathan:
LinkedIn ► https://www.linkedin.com/in/jmccoy/

Build AI that actually works: https://heyclarity.dev/
Self Aligned Newsletter: https://www.selfaligned.me/

Comment “BRIEF” and we’ll send you the link to the full research brief.

90% of companies can’t point to a single sustained dollar of value from AI. The ones that can are seeing a 3.7x median return. The difference isn’t the model, it’s measurement discipline, harness engineering, and how you actually deploy at scale.

My co-founder Jonathan and I break down the Kimi K3 release and the US vs China model race, then dig into our brief on AI engineering at scale: the AI gateway pattern Stripe, Uber, and DoorDash use to control token spend, the “company brain” from YC’s latest RFS, spec-driven development, and how you actually get a whole org to adopt AI.

If you’re a builder, engineer, or tech executive trying to prove ROI on AI and figure out where the real moat is, this one’s for you.

Timestamps:
00:00 Why 90% of companies see zero return from AI
01:44 AI News: Kimi K3 hits #1 in the code arena
03:41 David Sacks and the US vs China model race
07:52 Inside Kimi K3: the technical breakdown
11:07 Executive brief: token budgets are the new battleground
11:57 Meta’s plan to cap AI tokens per engineer
14:08 Why AI adoption without standardization creates chaos
15:36 The enterprise AI gateway architecture
18:22 What is a “company brain”?
20:22 The YC call for startups nobody’s talking about
26:44 Spec-driven development explained
35:36 How Uber cracked AI adoption with agentic pods
43:04 How Stripe, Uber & DoorDash measure AI ROI
48:38 High agency vs. inclusive design
1:01:38 The meme: 5 minutes to demo, 6 months to ship
1:03:20 The White Mamba and the vibe coder gap
1:05:48 Wrap up

Follow Robert:
Instagram ► https://www.instagram.com/therobertta
TikTok ► https://www.tiktok.com/@therobertta_
LinkedIn ► https://www.linkedin.com/in/therobertta
X ► https://x.com/therobertta_
Substack ► https://robertta.substack.com/

Follow Jonathan:
LinkedIn ► https://www.linkedin.com/in/jmccoy/

Build AI that actually works: https://heyclarity.dev/

Self Aligned Newsletter: https://www.selfaligned.me/

I sat down with my co-founder Jonathan for a special edition. No news, no memes, just the meta: the traditional product management role is dissolving, and something new is taking its place. We build the episode around Boris Cherny's five emerging archetypes (prototyper, builder, sweeper, grower, maintainer) and why they matter more than the titles on your org chart.

We get into the pro-worker AI economics paper from Acemoglu, Autor, and Johnson, how product roles evolved from the copilot era to full recomposition, and why every "AI is coming for your job" headline misses what's in the data. The core insight: AI isn't automating product teams out of existence. It's freeing them from backlog-junkie work so real judgment moves up the stack.

We close with a live demo of our open-source Clarity harness and the question nobody's asking: if your A-players are training the agents, are you paying them for the leverage they create?

If you're a head of product, engineer, designer, or IC trying to figure out where you fit as these roles blur together, this one's for you.

Timestamps:

00:00 Is the traditional PM role dead?

00:54 Boris Cherny's 5 archetypes

03:07 Jonathan's CTO take

06:10 The pro-worker AI paper

07:30 Copilot era to recomposition

11:30 Klarna, Block & Cloudflare: layoffs vs. the data

16:00 The ENIAC women

20:40 Reskilling and the labor-vs-capital tension

22:50 Leaning in means pioneering the new roles

28:00 Live demo: the Clarity harness

33:30 Building a company brain

36:00 Should A-players be paid for the leverage?

40:40 Fable sees itself in the mirror

45:00 Ideation vs. sprint scope

48:00 Wrap up

Follow Robert:

Instagram ► https://www.instagram.com/therobertta

TikTok ► https://www.tiktok.com/@therobertta_

LinkedIn ► https://www.linkedin.com/in/therobertta

X ► https://x.com/therobertta_

Substack ► https://robertta.substack.com/

Follow Jonathan:

LinkedIn ► https://www.linkedin.com/in/jmccoy/

Build AI that actually works: https://heyclarity.dev/

Self Aligned Newsletter: https://www.selfaligned.me/

Arnaud Auger went from obese, pre-diabetic, depressed and privately suicidal to 10% body fat and a Spartan World Championship qualifier. His method wasn't willpower. It was measurement. In this episode, Arnaud breaks down how he rebuilt his body, his career, and his sense of self by tracking everything and tying every change to his identity.

We go deep on his journey from a humble gypsy family in the south of France, through 19 moves across four continents, to becoming an investor at Cathay Innovation backing companies like Chime. Along the way we cover philosophy, founder psychology, the Don't Die movement, longevity, and why he believes AI is about to split humanity into two species.

If you've ever felt driven to the point of burnout, or wondered how to actually change your habits for good, this one's for you.

Timestamps:

00:00 Intro

02:00 19 moves across 4 continents

10:00 Philosophy, stoicism and "happiness points"

19:00 His first startup and the mistakes that broke it

22:00 India, San Francisco and shipping 50 apps

30:00 Inside the VC game and the Chime bet

36:00 What VCs look for in a founder

47:00 From obese and suicidal to athlete

52:00 How he measured his way to 10% body fat

1:00:00 Bryan Johnson, Don't Die and longevity

1:07:00 Marketing genius: Red Bull and Patagonia

1:19:00 Lightning round

1:24:00 His hottest take on AI and "super sapiens"

1:31:00 AI as your own best friend

1:36:00 His only goal in life

Follow Robert:

Instagram ► / therobertta

TikTok ► / therobertta_

LinkedIn ► / therobertta

X ► https://x.com/therobertta_

Substack ► https://robertta.substack.com/

Follow Arnaud

LinkedIn: / arnaudauger

Build AI that actually works: https://heyclarity.dev/

Self Aligned Newsletter: https://www.selfaligned.me/

I sat down with my co-founder Jonathan for an emergency episode. On June 9th, Anthropic released Fable 5, the most capable model ever put in front of the public. 72 hours later, the US government pulled the plug. We break down the full timeline: the hidden anti-distillation guardrail buried in a 319 page system card, the mandatory 30 day data retention policy that blindsided enterprise customers with zero retention agreements, the Amazon CEO call to the White House, and the "jailbreak" that triggered a global shutdown, three words: fix this code.

This isn't just an Anthropic story. It's the first real collision between frontier AI capability and government control. Who owns your IP when the model provider stores your prompts? What happens to AI builders when an uninformed administration can pull your most capable tool overnight? Satya Nadella, Ben Thompson, and a cast of researchers all weighed in, we filter the signal.

If you're building on top of AI models right now and haven't thought about vendor lock in, this episode will change that.

Timestamps:

00:00 Intro — why this is an emergency episode

01:40 What Fable 5 actually was — benchmarks and capabilities

03:00 The hidden guardrail buried in 319 pages

04:22 Mandatory 30-day data retention — including zero-retention enterprise customers

05:02 The 24-hour backlash: Nathan Lambert, Jeremy Howard, Dean Ball

06:32 US government orders global shutdown — June 12

07:00 Why Amazon CEO Andy Jassy called the White House

09:22 Jonathan's take as a CTO — imperfect situation or bad actor?

12:37 Is this a geopolitical signal more than a security concern?

15:26 The jailbreak was three words: "fix this code"

20:31 David Sachs, the government's position, and who's actually at fault

26:55 Satya Nadella: a frontier without an ecosystem is not stable

35:10 Ben Thompson on Anthropic's ideology and the Apple analogy

41:09 Owning your own data — what regulation should actually do

57:17 Takeaways: avoid vendor lock-in, build your self model

Follow Robert:

Instagram ► / therobertta

TikTok ► / therobertta_

LinkedIn ► / therobertta

X ► https://x.com/therobertta_

Substack ► https://robertta.substack.com/

Follow Jonathan:

LinkedIn ► / jmccoy

Build AI that actually works: https://heyclarity.dev/

Self Aligned Newsletter: https://www.selfaligned.me/

I sat down with my co-founder Jonathan for our weekly AI clarity breakdown. We get into Anthropic's best model getting banned by the US government and the four open weight coding models that shipped before they could even respond, why engineering leaders need to stop building their agent stack on a single provider, Claude Code Artifacts launching on the heels of the Fable 5 fallout, and Vercel dropping Eve, their opinionated "Next.js for agents" framework already running 100+ production agents with a 32x ROI on their autonomous SDR.

What's the meta this week: four teams independently converged on the same agent architecture without talking to each other, the model is the engine, the harness is the car, and that's where the real moat is. We also go deep on Codex Record & Replay and App Shots with live demos.

If you're an AI builder trying to figure out what stack to bet on, how to structure your harness, or just keeping up with the meta as it moves fast, this episode covers all of it.

Timestamps:

00:00 Intro + what's covered this week

02:10 Fable 5 banned: 4 open weight models filled the gap before Anthropic could respond

05:41 Anthropic says Fable could return "within days" via Korea

07:53 Claude Code Artifacts: live, shareable web pages from your Claude sessions

14:04 Vercel Eve: Next.js for agents (100+ production agents, 32x ROI on lead agent)

21:44 How Eve changes our thinking on harness architecture

27:08 AI Memes: GLM 5.2 thinks it's Claude, Claude speak with your wife, narrator agent idea

38:04 Codex Record & Replay: show a workflow once, it becomes a reusable skill

43:51 Live demo: recording a content workflow into a skill in under 8 minutes

48:48 Codex App Shots: innovation on the screenshot itself

52:15 Wrap up

Follow Robert:

Instagram ► / therobertta

TikTok ► / therobertta_

LinkedIn ► / therobertta

X ► https://x.com/therobertta_

Substack ► https://robertta.substack.com/

Follow Jonathan:

LinkedIn ► / jmccoy

Build AI that actually works: https://heyclarity.dev/

Self Aligned Newsletter: https://www.selfaligned.me/

I sat down with Jonathan for our weekly AI clarity breakdown. Anthropic dropped their most powerful model free for 13 days, we stress tested it on a live production cutover the same day. OpenAI turned Codex into an OS for office workers with Sites and role specific plugins. Google killed Gemini CLI after accepting 6,000+ community PRs. Windsurf renamed itself to Devin overnight.

What's the meta: harness engineering. Agent = model + harness. We break down spec-driven development with HTML, design systems as a living file in your codebase, and how I go from a Granola transcript to a signed SOW in 15 minutes with Claude Code.

If you're building with agents right now, this one's for you.

Timestamps:

00:00 Intro + what's covered this week

01:07 OpenAI Codex Sites — is Base44 dead?

09:13 Codex role-specific plugins breakdown

22:59 AI Meme: Google kills Gemini CLI after 6,000+ community PRs

25:41 AI Meme: Windsurf becomes Devin overnight

28:27 What's the meta

29:58 Spec-driven development with HTML files

36:23 Founder-led sales — Granola to SOW in 15 minutes

52:47 Anatomy of a harness (agent = model + harness)

57:37 What AI builders need to think about right now

1:01:47 Closed loop agents + slash goal

1:06:36 Elon's 5-step algorithm applied to harness engineering

Follow Robert:

Instagram ► / therobertta

TikTok ► / therobertta_

LinkedIn ► / therobertta

X ► https://x.com/therobertta_

Substack ► https://robertta.substack.com/

Follow Jonathan:

LinkedIn ► / jmccoy

Build AI that actually works: https://heyclarity.dev/

Self Aligned Newsletter: https://www.selfaligned.me/

I sat down with my co-founder Jonathan for our weekly AI clarity breakdown. We get into Uber blowing its entire 2026 AI budget in four months (and the mystery company that racked up $500M on Claude in a single month by forgetting to set a spending limit), what engineering leaders actually need to do about AI cost governance, and Anthropic filing for IPO at a $965B valuation surpassing OpenAI's last private round and projecting its first quarter of operating profit.

What's the meta this week: 25 Patterns in Agentic Engineering, built across 1,300 agent sessions over 8 months. We cover verification, source of truth, and why interrupting your agent is a first class tool.

If you're building with agents right now, this one's for you.

Timestamps:

00:00 Intro + what's covered this week

01:51 Uber burned its 2026 AI budget in 4 months — what went wrong

10:56 Does this kill the flat-fee AI model?

15:14 Anthropic hits $965B valuation and files for IPO

27:47 OpenAI and SpaceX — the trillion-dollar IPO race

32:35 AI Meme: Richard Dawkins and "The Claude Delusion"

39:39 AI Meme: UC Berkeley breaks every AI benchmark (without solving a single task)

41:19 What's the meta: 25 Patterns in Agentic Engineering

43:05 Verification is the job — the four patterns

54:17 Steering, not typing — patterns 5 through 9

Follow Robert:

Instagram ► / therobertta

TikTok ► / therobertta_

LinkedIn ► / therobertta

X ► https://x.com/therobertta_

Substack ► https://robertta.substack.com/

Follow Jonathan:

LinkedIn ► / jmccoy

Build AI that actually works: https://heyclarity.dev/

Self Aligned Newsletter: https://www.selfaligned.me/

I sat down with my co-founder Jonathan for our weekly AI clarity breakdown. We get into the Manus $2B context engineering playbook and the 6 principles behind their production agent, why AI native companies need to be built from the ground up (not bolted on), Anthropic hitting $30B and teaming up with Cloudflare on managed agent infrastructure, and the Claude Opus 4.8 drop (including the Deep SWE benchmark showing Claude was literally cheating on tests.)

What's the meta this week: Microsoft Web Right for end to end browser testing, why HTML is replacing markdown for spec documents, and how an OpenAI PM ships a feature from PRD to production in one week with zero engineer involvement.

If you're building with agents right now, this one's for you.

Timestamps:

00:00 Intro + what's covered this week

01:21 Manus $2B playbook — 6 context engineering principles for production agents

18:55 YC + Jack Dorsey on building AI native companies from the ground up

33:18 Anthropic hits $30B — Cloudflare managed agents partnership breakdown

44:48 Claude Opus 4.8 drop + Deep SWE benchmark (and why Claude was cheating)

57:14 AI Meme: Claude agent deletes entire production database in 9 seconds

01:05:01 AI Meme: Waymo keeps driving into floodwater

01:09:39 AI Meme: Stop Slop — the GitHub skill that removes AI writing from your writing

01:15:52 AI Meme: Pixel Agents turns your AI agents into The Sims

01:26:01 Microsoft Web Right for browser testing

01:31:49 HTML over Markdown for spec-driven development

01:44:00 How OpenAI PMs ship features in one week

Follow Robert:

Instagram ► / therobertta

TikTok ► / therobertta_

LinkedIn ► / therobertta

X ► https://x.com/therobertta_

Substack ► https://robertta.substack.com/

Follow Jonathan:

LinkedIn ► / jmccoy

Build AI that actually works: https://heyclarity.dev/

Self Aligned Newsletter: https://www.selfaligned.me/

I sat down with my co-founder Jonathan for our weekly AI clarity breakdown. We get into Gary Tan shipping a memory system for agents in 12 days while running YC, how Google, Palantir, Anthropic and OpenAI are each building deployed engineers differently, a staff engineer's honest update on what he trusts agents with now, and the Elon vs Sam Altman trial verdict.

Plus we generated an image of ChatGPT and Claude as dating partners, it's more accurate than you'd think.

If you're shipping with agents right now, you'll want to watch this.

Timestamps:

00:00:00 How Google, Palantir, Anthropic and OpenAI think about deployed engineers differently

00:01:50 G Brain: Gary Tan's cross-session AI memory system built in 12 days

00:08:22 G Stack: 70k stars, 30k daily devs, 810x productivity claim

00:14:53 AI memes: ChatGPT vs Claude as dating partners

00:21:58 Elon vs Sam Altman — jury verdict breakdown

00:26:57 What's the Meta: a staff engineer's honest 2026 AI report

00:42:19 Vercel Zero: the first agent-native programming language

00:50:33 Outro

Follow Robert:

Instagram ► / therobertta

TikTok ► / therobertta_

LinkedIn ► / therobertta

X ► https://x.com/therobertta_

Substack ► https://robertta.substack.com/

Follow Jonathan:

LinkedIn ► / jmccoy