Company · Frontier labs
Anthropic revenue, valuation & funding
Anthropic was valued at $965B in its Series H in May 2026, and it said its run-rate revenue crossed $47B that month. It's still private. It filed in October 2026 to list on the Nasdaq under the symbol ANT.
At a glance
- Valuation: $965B after its $65B Series H in May 2026, its latest priced round.
- Revenue: a run rate above $47B in May 2026, the company said. Sacra, reading its S-1, puts revenue for the first nine months of 2026 at $33.242B.
- Status: private, and its shares don't trade yet. It filed an S-1 in October 2026 to list on Nasdaq as ANT.
- Founded: January 2021, by former OpenAI employees including Dario and Daniela Amodei.
- CEO: Dario Amodei, a co-founder and board member.
- Headquarters: 500 Howard Street, San Francisco.
- Employees: approximately 5,300 as of September 30, 2026, per its S-1.
- Investors: Altimeter Capital, Dragoneer, Greenoaks, Sequoia Capital, GIC, Coatue, ICONIQ, Lightspeed and Amazon, among others.
- Products: the Claude models, including Opus 5.5 and Sonnet 5.5, the Claude apps, Claude Code, Cowork and the Claude API.
- On Real Alpha: 114 moments of its people talking, from 17 interviews.
Is Anthropic public? Can you buy its stock?
Not yet. Anthropic is private, and its shares don't trade on any exchange. It has applied to list its Class A common stock on the Nasdaq Global Select Market under the symbol ANT, according to its Form S-1. The filing left the price range and share count blank.
Anthropic said on June 1, 2026 that it had filed confidential IPO papers with the SEC, according to Wikipedia. It filed the registration statement publicly on October 26, 2026. A first amendment followed on November 2, with financial statements through September 30 (Sacra).
Until a listing, its shares are held by its founders, staff and investors. In February 2026 an employee tender offer of $5 billion to $6 billion let current and former staff sell shares at about $350 billion, Sacra estimates. This page reports facts and gives no investment advice.
Valuation, round by round
Its valuation went from $61.5B in March 2025 to $965B in May 2026, more than fifteen times in just over a year.
| Date | Valued at | Raised | Led by | Source |
|---|---|---|---|---|
| May 2026 | $965B | $65B | Altimeter Capital, Dragoneer, Greenoaks, and Sequoia Capital | Anthropic |
| Feb 2026 | $380B | $30B | GIC and Coatue | Anthropic |
| Sep 2025 | $183B | $13B | ICONIQ | Anthropic |
| Mar 2025 | $61.5B | $3.5B | Lightspeed Venture Partners | Anthropic |
A $124M Series A in May 2021, led by Jaan Tallinn, appears only on data sites and isn't confirmed by a primary report. Amounts and valuations for the Series B, C and D weren't found, and the Series D is dated only to early 2024.
Revenue
Anthropic said its run-rate revenue crossed $47 billion in May 2026, up from $14 billion in February and over $5 billion in August 2025. Sacra estimates it reached $65B in July 2026.
| Date | Figure | Basis | Source |
|---|---|---|---|
| Jul 2026 | $65B | Estimate | Sacra |
| May 2026 | $47B | Company | Anthropic |
| Feb 2026 | $14B | Company | Anthropic |
| Sep 2025 | $5B+ | Company | Anthropic |
A run rate is a month's revenue multiplied by twelve, which differs from audited revenue. Sacra's reading of the S-1 puts revenue at $4.585B for all of 2025 and $33.242B for the first nine months of 2026.
How Anthropic makes money
Anthropic earns money from usage-based access to its Claude models, from subscriptions for people and businesses, and from enterprise agreements, Sacra says. Dario Amodei has said the company puts business customers first.
Compute is its largest commitment. Its November 2 IPO amendment disclosed approximately $895.4B in purchase commitments and lease agreements, mostly for compute and infrastructure, per Sacra. Relaying The Wall Street Journal's review of the draft prospectus, Sacra estimates approximately $450B in cloud-computing agreements over seven to ten years, most of them noncancelable.
Sacra estimates recognized revenue at $16.3B in the first half of 2026, up from $1.1B a year earlier. Sacra estimates an $80.7B net loss for the same half, against $13B a year earlier, with about $72B in non-cash accounting charges. Sacra estimates the adjusted operating margin, which leaves out stock-based compensation and non-cash charitable contributions, at 15% in Q3 2026, down from 19% in Q2.
- Subscriptions. Claude Pro costs $20 a month billed monthly, Sacra estimates. Max plans cost $100 or $200 a month for more usage, by the same estimate. Sacra
- Team plans. Standard seats cost $20 per user per month billed annually, or $25 monthly, Sacra estimates. Premium seats cost $100 billed annually, or $125 monthly, by the same estimate. Sacra
- API usage. Per million tokens, Sonnet 5.5 costs $2 for input and $10 for output, Sacra estimates. Opus 5.5 costs $4 and $20, and Fable 5.1 costs $10 and $50, by the same estimate. Sacra
- Partners. Anthropic put $100 million into the Claude Partner Network in March 2026. Anthropic's blog
- Snowflake. A multi-year $200M partnership pairs model access with joint enterprise sales, Sacra says. Sacra
From Anthropic's interviews
- 29:46 Each model can make money while the company spends at a lossDario Amodei · Big Technology Podcast · Jul 2025
- 58:45 AI lab profits hinge on compute allocation and demand forecastsDario Amodei · Dwarkesh Podcast · Feb 2026
- 46:49 Compute commitments must price in a one-year demand missDario Amodei · Dwarkesh Podcast · Feb 2026
- 23:34 Enterprise customers may value model improvements more than consumersDario Amodei · Big Technology Podcast · Jul 2025
Products
- Claude (released March 2023, per Wikipedia). The first version finished training in summer 2022, and Anthropic held it back for safety testing. Wikipedia
- Opus 5.5 and Sonnet 5.5. Sacra describes them as the current lineup. Opus 5.5 is for complex coding and professional work, and Sonnet 5.5 for faster, lower-cost tasks. Sacra
- Claude Code (general availability May 2025). A coding agent that started as a terminal prototype built by Boris Cherny. Wikipedia
- Cowork. An app for people who don't write code, grown from how non-engineers used Claude Code. Boris Cherny said a small team prototyped it in about 10 days.
- Claude API and SDKs. Anthropic said on May 18, 2026 that it was acquiring Stainless, a maker of SDK and MCP server tooling. Terms were not disclosed. Anthropic's blog
From Anthropic's interviews
- 2:33 Claude Code was built before the model was readyBoris Cherny · Training Data · May 2026
- 2:20 Claude Code began as a lightweight terminal prototypeBoris Cherny · AI & I · Oct 2025
- 1:05:11 Cowork was built quickly, with safety as the hard partBoris Cherny · The Pragmatic Engineer · Mar 2026
- 1:27:18 Claude Code emerged from Anthropic's internal research workflowDario Amodei · Dwarkesh Podcast · Feb 2026
Customers
Dario Amodei has said enterprise adoption is fast, though procurement and rollout still take time.
Customers paying Anthropic $100,000 or more over the trailing twelve months grew from more than 950 to more than 11,900 in the year to September 30, 2026, its S-1 says.
Two customers made up approximately 20% of first-half 2026 revenue, per Sacra's reading of the prospectus.
Inside the company, Cat Wu said in October 2025 that 70% to 80% of technical staff used Claude Code every day.
From Anthropic's interviews
- 20:23 Anthropic reported rapid revenue growth and enterprise adoptionDario Amodei · Dwarkesh Podcast · Feb 2026
- 7:57 Anthropic's internal dogfooding gives Claude Code rapid feedbackCat Wu · AI & I · Oct 2025
- 21:07 Claude Code coincided with a 200% rise in PRs per engineerBoris Cherny · Lenny's Podcast · Feb 2026
- 31:56 Early-access model testing depends on customers with robust evalsKatelyn Lesse · Latent Space · Oct 2026
Leadership
Anthropic had approximately 5,300 employees as of September 30, 2026, according to its S-1.
- Dario Amodei, co-founder, CEO and a member of the board. He earlier led OpenAI's GPT-2 and GPT-3 research. SEC (Form S-1)
- Daniela Amodei, co-founder, President and Chairperson of the board. She was earlier VP of Safety and Policy at OpenAI. SEC (Form S-1)
- Krishna Rao, chief financial officer. Anthropic's blog
- Ami Vora, chief product officer, per Wikipedia. Wikipedia
- Mariano-Florentino Cuéllar, chief global affairs officer, per Wikipedia. Wikipedia
- Jared Kaplan, co-founder and chief science officer. He co-authored the scaling laws paper on neural language models.
- Jack Clark, co-founder. He leads The Anthropic Institute, which studies the effects of frontier AI on society.
- Chris Olah, co-founder. He leads its interpretability research.
- Mike Krieger, co-leads Anthropic Labs, which incubates experimental Claude products. He co-founded Instagram and was Anthropic's chief product officer.
- Boris Cherny, created Claude Code and leads its team.
- Cat Wu, head of product for Claude Code and Cowork.
- Katelyn Lesse, head of platform engineering. Anthropic's blog
- Angela Jiang, led product for the Claude Platform after leading API at OpenAI.
- Amanda Askell, leads the personality alignment team that shapes Claude's character.
- Jan Leike, alignment researcher. He joined in 2024 after co-leading OpenAI's superalignment project.
- Andrej Karpathy, joined the pretraining team. He founded Eureka Labs and led Tesla's Autopilot AI.
- Emmanuel Ameisen, interpretability researcher tracing how Claude reasons.
From Anthropic's interviews
- 2:17:03 Amodei uses direct communication to maintain Anthropic’s cultureDario Amodei · Dwarkesh Podcast · Feb 2026
- 1:38:10 Talent density matters more than headcountDario Amodei · Lex Fridman Podcast · Nov 2024
- 16:13 Anthropic kept fixed pay levels during Meta's recruiting pushDario Amodei · Big Technology Podcast · Jul 2025
- 17:18 Anthropic's workflow advantage is organizational, not model accessBoris Cherny · Training Data · May 2026
Who owns Anthropic?
Its founders, staff and investors. Under the S-1, a Class F share held by a founders' LLC would carry 50.1% of total voting power. That would make Anthropic a controlled company.
- Altimeter Capital, Dragoneer, Greenoaks and Sequoia Capital led the $65 billion Series H in May 2026. Anthropic
- Amazon put $5 billion into the Series H, part of $15 billion that hyperscalers had committed earlier. Anthropic
- GIC and Coatue led the $30 billion Series G in February 2026. Anthropic
- ICONIQ led the $13 billion Series F in September 2025, with Fidelity and Lightspeed as co-leads. Anthropic
- Lightspeed Venture Partners led the $3.5 billion Series E in March 2025. Anthropic
- Microsoft and Nvidia committed in November 2025 to invest up to $5 billion and up to $10 billion, Sacra says. Sacra
- Menlo Ventures led the Series D in early 2024. Revenue Memo
- Spark Capital led the Series C in May 2023, and its partner Yasmin Razavi joined the board. Revenue Memo
- FTX and Alameda supplied most of the Series B in April 2022, which Sam Bankman-Fried led. Revenue Memo
Competitors
Anthropic competes with other labs in foundation models, developer APIs and coding agents.
- OpenAI, its most direct competitor, Sacra says. Claude and Claude Code compete with GPT models and Codex. Sacra
- Google, which competes across models, cloud distribution and productivity software, Sacra says. Sacra
- Chinese labs. In February 2026 Anthropic accused DeepSeek, Moonshot AI and MiniMax of distillation attacks. It alleged over 16 million chats with Claude from around 24,000 fraudulent accounts. Wikipedia
What its people said about the competition
- 36:35 Open weights do not make AI models work like open-source softwareDario Amodei · Big Technology Podcast · Jul 2025
- 9:41 Anthropic makes reasoning harder to distill without breaking API useKatelyn Lesse · Latent Space · Oct 2026
Timeline and latest news
From its founding to the latest reports, newest first.
- Nov 2, 2026Latest
Filed a first S-1 amendment disclosing about $895.4B in purchase commitments and leases, Sacra estimates. Sacra - Oct 26, 2026Latest
Filed its Form S-1 publicly with the SEC. SEC (Form S-1) - Sep 25, 2026Latest
An appeals court upheld the Pentagon's supply chain risk designation. ABC News - Jul 20, 2026
A judge gave final approval to its $1.5 billion copyright settlement with authors. TechCrunch - Jun 1, 2026
Announced it had filed confidential IPO papers with the SEC. Wikipedia - May 28, 2026
Raised $65 billion at a $965 billion valuation in its Series H. Anthropic - May 18, 2026
Announced it was acquiring Stainless, a maker of SDK tooling. Anthropic's blog - Mar 2026
The Department of Defense designated it a supply chain risk. Wikipedia - Feb 23, 2026
Accused DeepSeek, Moonshot AI and MiniMax of distillation attacks on Claude. Wikipedia - Feb 12, 2026
Raised $30 billion at a $380 billion valuation, led by GIC and Coatue. Anthropic - Oct 2025
Announced a cloud deal for access to up to one million Google TPUs. Wikipedia - Sep 2, 2025
Raised $13 billion at a $183 billion valuation, led by ICONIQ. Anthropic - May 2025
Claude Code moved from research preview to general availability. Wikipedia - Mar 3, 2025
Raised $3.5 billion at a $61.5 billion valuation, led by Lightspeed. Anthropic - Mar 2023
Released the first version of Claude. Wikipedia - Jan 2021
Founded by former OpenAI employees, including Dario and Daniela Amodei. Wikipedia
Legal matters
Anthropic has settled a copyright class action and is challenging a Pentagon designation in court.
- Authors' copyright class action (filed August 2024). A federal judge gave final approval on July 20, 2026 to a $1.5 billion settlement. The case is closed, with payments to authors and publishers. TechCrunch
- Pentagon designation (March 2026). The Department of Defense named Anthropic a supply chain risk and barred military contractors from working with it. It followed Anthropic's refusal to drop limits on surveillance and autonomous-weapons use. Wikipedia
- Government phase-out. A federal judge's preliminary injunction blocked the move to phase Anthropic products out of government use, Wikipedia says. The judge called it First Amendment retaliation. Wikipedia
- Appeal. On September 25, 2026 the D.C. Circuit upheld the Pentagon's designation. Anthropic is considering further review. ABC News
In their own words, and what followed
What people said about the business, with the minute they said it, and what was reported after.
Dario Amodei on Dwarkesh Podcast · Feb 13, 2026
“there's this bizarre 10x per year growth in revenue that we've seen”
20:23 Anthropic reported rapid revenue growth and enterprise adoption
Reported after
With its Series H in May 2026, Anthropic said its run-rate revenue had crossed $47 billion. Anthropic
Dario Amodei on Dwarkesh Podcast · Feb 13, 2026
“If I'm just off by a year in that rate of growth”
46:49 Compute commitments must price in a one-year demand miss
Reported after
Its November 2 IPO amendment disclosed approximately $895.4B in purchase commitments and lease agreements, mostly for compute, Sacra reported. Sacra
Dario Amodei on Dwarkesh Podcast · Feb 13, 2026
“if every year we predict exactly what the demand is going to be, we'll be profitable every year.”
58:45 AI lab profits hinge on compute allocation and demand forecasts
Reported after
Sacra, reading the IPO filing, put the adjusted operating margin at 15% in Q3 2026, down from 19% in Q2. Sacra
Dario Amodei on Big Technology Podcast · Jul 30, 2025
“the enterprise use of AI is going to be greater even than the consumer use of AI”
23:34 Enterprise customers may value model improvements more than consumers
Reported after
Its S-1 said customers paying $100,000 or more a year grew from more than 950 to more than 11,900 by September 2026. SEC (Form S-1)
What others say about Anthropic
Moments where people outside the company talk about it.
David Sacks · All-In · Oct 30, 2026
“Anthropic has been an extraordinarily capital-efficient business.”
Ed Zitron · Newcomer Pod · Jun 8, 2026
“Anthropic has never defined what they mean by run rate.”
Dan Primack · Big Technology Podcast · Oct 26, 2026
“a bet that most people weren't making, you know, a couple of years ago to start there”
14:04 Anthropic’s B2B bet helped diversify its customer base 3 min
John Foley · Prof G Markets · Nov 3, 2026
“it is almost impossible to predict with any certainty where these revenue numbers are going to go.”
2:35 Anthropic’s revenue forecasts are difficult to underwrite 3 min
John Coogan · TBPN · Oct 27, 2026
“Anthropic's IPO includes a private line of succession”
36:07 Anthropic founders can appoint successors to voting control 3 min
Questions people ask
Is Anthropic a public company? Can I buy Anthropic stock?
Not yet. Anthropic is private and its shares don't trade on any exchange. It has applied to list on the Nasdaq Global Select Market under the symbol ANT.
When will Anthropic IPO?
It filed its S-1 publicly on October 26, 2026 and a first amendment on November 2. The filing left the price range and share count blank.
Who owns Anthropic?
Its founders, staff and investors, including Altimeter Capital, Dragoneer, Greenoaks, Sequoia Capital, GIC, Coatue, ICONIQ, Lightspeed and Amazon. Under the S-1, a founders' LLC would hold 50.1% of voting power.
Who is the CEO of Anthropic?
Dario Amodei, a co-founder who also sits on the board. He earlier led OpenAI's GPT-2 and GPT-3 research. 2:17:03
Who founded Anthropic?
Former OpenAI employees founded it in January 2021. Wikipedia lists Dario Amodei, Daniela Amodei, Jared Kaplan, Jack Clark, Chris Olah, Ben Mann, Sam McCandlish and Tom Brown.
How much is Anthropic worth?
$965 billion after its $65 billion Series H in May 2026, its latest priced round.
How much revenue does Anthropic make?
Anthropic said its run-rate revenue crossed $47 billion in May 2026. Sacra, reading the S-1, puts revenue for the first nine months of 2026 at $33.242B.
Is Anthropic profitable?
It posted a net loss in the first half of 2026. Sacra counts $80.7B, with about $72B in non-cash charges. Its adjusted operating margin was 15% in Q3, Sacra estimates. 58:45
How does Anthropic make money?
It sells usage of its Claude models through its API, subscriptions from $20 to $200 a month, Team plans and enterprise agreements. 29:46
How many customers does Anthropic have?
More than 11,900 customers paid it $100,000 or more in the year to September 30, 2026, up from more than 950, its S-1 says. 20:23
Where is Anthropic headquartered?
500 Howard Street in San Francisco.
Who are Anthropic's competitors?
OpenAI is its most direct rival, Sacra says. Google competes across models, cloud and productivity software.
Every moment from Anthropic's people
114 moments from 17 interviews on 13 shows, newest first. Each one plays from the minute it starts.
Latent Space · Oct 27, 2026 · 7 moments
How Anthropic ships the #1 LLM in the World — with Katelyn Lesse and Angela Jiang
- 56:02 Claude helped cut Messages API CPU use by up to 50% 3 min
- 1:03:18 Separating the agent from its sandbox limits security risks 5 min
- 9:41 Anthropic makes reasoning harder to distill without breaking API use 3 min
- 19:08 Model launches now include policy and authorization work 3 min
- 31:56 Early-access model testing depends on customers with robust evals 4 min
- 1:20:59 Humans own system design for Claude-written code 4 min
- 1:43:26 Parallel agents need shared context and asynchronous work 3 min
Lenny's Podcast · Oct 18, 2026 · 3 moments
A mind-expanding conversation on how models actually work | Emmanuel Ameisen (Anthropic)
Lightcone · Jul 27, 2026 · 6 moments
Boris Cherny: We Cut 80% of Claude Code’s Prompt
- 14:31 Give models harder tasks to reveal hidden capabilities 5 min
- 19:29 Verify hard tasks instead of over-specifying them 5 min
- 24:51 Agent workflows can automate recurring codebase maintenance 6 min
- 7:04 Add prompt instructions only after repeated failures 4 min
- 3:07 Opus 5 lets Claude Code remove instructions it no longer needs 4 min
- 10:18 Claude Code gave models write access to whole files 4 min
Odd Lots · Jul 20, 2026 · 1 moment
The Creator of Claude Code on The Hottest Piece of Software in the World | Odd Lots
Show 13 more interviews
Training Data · May 4, 2026 · 6 moments
Anthropic's Boris Cherny: Why Coding Is Solved, and What Comes Next
- 5:04 Coding is solved for the code Boris writes 4 min
- 10:22 AI weakens some software moats but strengthens others 3 min
- 14:57 Software could become as ordinary as reading and writing 2 min
- 2:33 Claude Code was built before the model was ready 3 min
- 13:03 Product craft still matters as models improve 2 min
- 17:18 Anthropic's workflow advantage is organizational, not model access 2 min
Training Data · Apr 29, 2026 · 6 moments
Andrej Karpathy: From Vibe Coding to Agentic Engineering w/ Stephanie Zhan
- 2:24 Software 3.0 makes some apps unnecessary 5 min
- 9:29 Verifiability helps explain where AI excels 4 min
- 19:22 Engineers still need to own judgment and system design 4 min
- 15:42 Agentic engineering raises output without lowering the quality bar 4 min
- 25:15 Agent-first services should let agents handle setup and deployment 2 min
- 27:40 You can outsource thinking but not understanding 2 min
Newcomer Pod · Apr 20, 2026 · 8 moments
Amanda Askell on AI Consciousness, Claude & Silicon Valley’s Biggest Fear
- 15:03 Why AI systems need judgment, not total obedience 4 min
- 24:24 AI self-reports are weak evidence of consciousness 5 min
- 29:12 Uncertainty about AI consciousness calls for basic respect 3 min
- 47:46 Verified work contexts can unlock dual-use assistance 5 min
- 39:00 Constitution training draws on existing model knowledge 3 min
- 31:55 AI could expand scarce expertise for rare diseases 4 min
- 9:23 Public constitutions make subjective model grading contestable 2 min
- 5:00 Claude once ended a late-night task like a human colleague 3 min
No Priors · Mar 20, 2026 · 8 moments
Skill Issue: Andrej Karpathy on Code Agents, AutoResearch, and the Loopy Era of AI
- 16:21 Autonomous research removes the researcher as the bottleneck 7 min
- 37:47 Cheaper software could increase demand for engineering 6 min
- 48:48 Open models can balance the power of frontier labs 6 min
- 33:02 Distributed agents can improve models if work is verifiable 4 min
- 54:19 AI will move from digital work to the physical interface 6 min
- 44:29 Working inside a frontier lab can constrain independent judgment 4 min
- 9:10 Agents could replace bespoke apps with exposed APIs 6 min
- 1:01:29 Libraries should document their code for agents 5 min
The Pragmatic Engineer · Mar 4, 2026 · 8 moments
Building Claude Code with Boris Cherny
- 24:03 Claude Code grew from letting the model use tools 3 min
- 40:26 Repeated review comments can become lint rules 6 min
- 1:05:11 Cowork was built quickly, with safety as the hard part 6 min
- 32:54 A clear plan lets parallel agents implement code 8 min
- 49:39 Agentic search beat a local vector index 3 min
- 17:00 Code quality contributed double-digit productivity gains at Meta 3 min
- 1:14:42 Uncorrelated context windows improve complex agent tasks 3 min
- 1:25:45 The printing press multiplied output before literacy spread 4 min
Lenny's Podcast · Feb 19, 2026 · 8 moments
Head of Claude Code: What happens after coding is solved | Boris Cherny
- 23:55 Give engineers tokens before optimizing costs 4 min
- 47:20 User workarounds reveal what product to build next 5 min
- 1:03:15 Build AI products around the model's expanding abilities 5 min
- 8:41 The terminal was the only interface that could keep pace 5 min
- 21:07 Claude Code coincided with a 200% rise in PRs per engineer 3 min
- 32:06 AI could make software creation as widespread as literacy 4 min
- 54:21 Agent safety needs real-world testing beyond labs and evals 5 min
- 1:08:42 Plan mode can make Claude Code tasks more reliable 3 min
Dwarkesh Podcast · Feb 13, 2026 · 8 moments
Dario Amodei — “We are near the end of the exponential”
- 46:49 Compute commitments must price in a one-year demand miss 7 min
- 58:45 AI lab profits hinge on compute allocation and demand forecasts 8 min
- 1:27:18 Claude Code emerged from Anthropic's internal research workflow 3 min
- 20:23 Anthropic reported rapid revenue growth and enterprise adoption 7 min
- 36:56 Anthropic estimates coding models add 15–20% to productivity 6 min
- 1:37:15 Amodei favors federal AI standards over a state-law moratorium 7 min
- 1:47:36 AI could give democracies leverage in setting global rules 7 min
- 2:17:03 Amodei uses direct communication to maintain Anthropic’s culture 5 min
The Peterman Pod · Dec 15, 2025 · 8 moments
Boris Cherny (Creator of Claude Code) On What Grew His Career And Building at Anthropic
- 1:10:24 Build for the model that will exist six months from now 2 min
- 29:37 Use competing designs to make large projects easier to scope 6 min
- 44:33 Turn repeated engineering pain into leverage for the team 6 min
- 7:01 Cross-org projects need a shared goal 3 min
- 18:55 Volunteering for a platform rewrite can earn influence 5 min
- 35:31 A large data-model migration can be worth reversing 6 min
- 54:13 Trust helped Boris move Instagram from Python to Hack 6 min
- 1:13:21 AI-generated code should meet the same review bar 3 min
AI & I · Oct 29, 2025 · 6 moments
The Secrets of Claude Code From the Engineers Who Built It
- 26:11 Turn past code and mistakes into compounding leverage 7 min
- 21:43 Use parallel agents to migrate code and filter false positives 5 min
- 2:20 Claude Code began as a lightweight terminal prototype 5 min
- 40:57 Agentic search avoids stale indexes and simplifies deployment 4 min
- 45:01 The next coding models may run autonomously for days 3 min
- 7:57 Anthropic's internal dogfooding gives Claude Code rapid feedback 6 min
Dwarkesh Podcast · Oct 17, 2025 · 8 moments
Andrej Karpathy — “We’re summoning ghosts, not building animals”
- 40:53 Outcome-based RL spreads noisy credit across a whole trace 8 min
- 1:22:15 AI may preserve the long-run 2% growth trend 5 min
- 1:59:10 A good tutor keeps each lesson just beyond the student's reach 7 min
- 8:17 Pretraining creates digital agents, not evolved animals 7 min
- 30:27 Karpathy uses autocomplete for novel code and agents for boilerplate 5 min
- 51:52 Synthetic training data can narrow a model's range of thought 8 min
- 1:43:43 Self-driving progress requires repeated rounds of reliability work 8 min
- 2:15:20 Teach by exposing the simplest useful model first 4 min
Big Technology Podcast · Jul 30, 2025 · 8 moments
Anthropic CEO Dario Amodei: AI's Potential, OpenAI Rivalry, GenAI Business, Doomerism
- 3:28 AI timelines hinge on an exponential, not an AGI label 6 min
- 29:46 Each model can make money while the company spends at a loss 7 min
- 58:51 Safety needs to advance alongside increasingly capable AI 7 min
- 11:36 Context windows can cover some continual-learning gaps 4 min
- 16:13 Anthropic kept fixed pay levels during Meta's recruiting push 4 min
- 23:34 Enterprise customers may value model improvements more than consumers 3 min
- 36:35 Open weights do not make AI models work like open-source software 4 min
- 53:54 Anthropic publishes safety policies for rivals to adopt 3 min
Lightcone · Jun 19, 2025 · 7 moments
Andrej Karpathy: Software Is Changing (Again)
- 18:14 Build AI apps around human review and adjustable autonomy 6 min
- 33:22 Make software and documentation easier for agents to use 5 min
- 7:51 LLMs look more like operating systems than utilities 7 min
- 26:08 A flawless self-driving demo did not mean autonomy was near 3 min
- 30:54 Shipping an AI-built app took longer than writing its code 2 min
- 3:51 Neural nets gradually replaced Autopilot's hand-written code 3 min
- 14:47 LLMs do not build organizational context on their own 4 min
Lex Fridman Podcast · Nov 11, 2024 · 8 moments
Dario Amodei: Anthropic CEO on Claude, AGI & the Future of AI & Humanity | Lex Fridman Podcast #452
- 1:38:10 Talent density matters more than headcount 4 min
- 2:05:49 Physical and institutional limits slow AI-driven change 6 min
- 4:40:10 Superposition lets networks represent more features than dimensions 5 min
- 15:32 Frontier compute budgets could reach $100 billion by 2027 5 min
- 58:49 Anthropic ties safety measures to capability thresholds 8 min
- 1:06:19 ASL-4 tests may need evidence beyond model answers 3 min
- 3:05:54 Prompt iteration matters most when optimizing the final few percent 6 min
- 5:00:54 Sparse features can expose deception-related model behavior 4 min
People on Real Alpha
- Dario Amodei, Co-founder & CEO, Anthropic · 21 moments
- Daniela Amodei, Co-founder & President, Anthropic
- Andrej Karpathy, Anthropic; founder, Eureka Labs · 29 moments
- Mike Krieger, Anthropic; co-founder, Instagram
- Boris Cherny, Creator of Claude Code, Anthropic · 42 moments
- Jack Clark, Co-founder, Anthropic
- Jared Kaplan, Co-founder & Chief Science Officer, Anthropic
- Chris Olah, Co-founder, Anthropic · 2 moments
- Jan Leike, Alignment researcher, Anthropic
- Emmanuel Ameisen, Interpretability researcher, Anthropic · 3 moments
- Cat Wu, Head of product, Anthropic · 1 moment
- Amanda Askell, Head of personality alignment, Anthropic · 9 moments
- Katelyn Lesse, Head of Platform Engineering, Anthropic · 5 moments
- Angela Jiang, Head of product, Claude Platform, Anthropic · 2 moments
Other companies in Frontier labs
OpenAI · Google DeepMind · Meta · SpaceXAI · Microsoft AI · Mistral AI · DeepSeek · Safe Superintelligence · Thinking Machines Lab · Cohere · Moonshot AI · Reflection AI
How this page is sourced
Each figure links to the filing, company statement or news report it came from, with its date. Run rates are what reporters were told, not audited results; estimates and unconfirmed reports are marked. Quotes are what each person said on the record, with the minute they said it. Spotted something that's off? .
Last checked Nov 7, 2026