Anthropic's Boris Cherny: Why Coding Is Solved, and What Comes Next

Punti chiave

1. Data / fonte. intervista video di Sequoia Capital (YouTube), URL https://www.youtube.com/watch?v=SlGRN8jh2RI. Data esatta non visibile nella trascrizione; riferimenti interni (Opus 4.7 come modello attuale, "we also just launched routines") la collocano a fine aprile / inizio maggio 2026. Data di aggiunta alla cartella di veille: 2026-05-05.
2. Relatore. Boris Cherny, creatore di Claude Code, Anthropic. Intervistatrice: Lauren Reader (Sequoia). Introduzione: Asia (Sequoia).
3. Il background di Cherny rivelato.
4. Ex allievo di YC, primo dipendente di un'azienda YC, diverse startup precedenti.
5. Autore di un libro, "Programming TypeScript."
6. Non ha scritto una sola riga di codice nel 2026. (dichiarazione fatta a Sequoia prima dell'intervista).
7. Aneddoto delle scuole medie: ha scritto una guida BASIC per la TI-83 Plus, "still on the internet, extremely embarrassing."
8. Apertura comica.
9. "Show of hands, who here uses Claude Code?" → maggioranza.
10. "Show of hands, who here has Claude code psychosis?" → risate.
11. "My team lovingly says I have Claude code psychosis, which may or may not be true." ### Storia di Claude Code (nuovo racconto)
12. Origine. Cherny entra in Anthropic Labs alla fine del 2024 — l'incubatore interno di Anthropic, un team di poche persone.
13. Output di Anthropic Labs. Claude Code, MCP, app desktop. Il team è stato sciolto una volta compiuta la missione.
14. Round 2 attuale. team riformato, guidato da Mike Krieger (Chief Product Officer di Anthropic, co-fondatore ex-Instagram).
15. Perché Cherny è partito dal coding. "product overhang" — "the model can do all the stuff that no product has yet captured."
16. Stato dell'arte del coding a fine 2024. "type ahead" (apri l'IDE, premi tab, completa una riga). È ciò che Sonnet 3.5 aveva sbloccato.
17. La scommessa. andare oltre — "the model was almost ready for the next big step. So we don't have to do type ahead anymore, we can just have the agent write all of the code."
18. Primi 6 mesi. "it just really didn't work. It was like not very good. It was barely usable. I used it for maybe 10% of my code."
19. Prima release di Claude Code. "It was not a hit. There's a lot of people that used it, but it did not have this exponential growth that it has today."
20. Punto di svolta. Opus 4 nel maggio 2025. "That's like when the exponential growth started, and then it kind of inflected with every model release. It started with Opus 4, then 4.5, then 4.6, now 4.7. It just kind of keeps inflecting."
21. Scommessa pre-PMF deliberata. "essentially we were trying to build this thing that was pre-PMF, and we knew that it wouldn't have PMF for 6 months because we were building for the next model." ### "Coding is solved"
22. Sondaggio a mano alzata:
23. "Who writes 100% of their code by hand?" → alcuni
24. "Who writes 100% using an agent like Claude Code?" → alcuni
25. "Who's somewhere in between?" → maggioranza ("like 50% solved")
26. Cherny stesso: 100%. .
27. Codebase di Claude Code.
28. "It leaked, so people know."
29. TypeScript + React. — scelti perché "very on distribution for the model" (all'epoca il modello era forte in queste tecnologie).
30. "Pretty simple. There's nothing really complicated."
31. Il modello ha scritto il 100% del codice da ottobre/novembre 2025. .
32. "For me today, the model writes 100% of my code. I write somewhere usually a few dozen PRs every day."
33. "There was a day last week I did like 150 PRs in a day. That was a record."
34. Limiti riconosciuti. "this is not the case everywhere. There's very big complicated codebases. There's weird languages the model isn't good at yet. Usually the answer is just wait for the next model." ### Configurazione personale attuale (aggiornamenti)
35. iOS prima di tutto. "most of my work I do from my phone."
36. App Claude → scheda codice. "a little code tab on the left-hand side, I just have a bunch of sessions going."
37. 5-10 sessioni. aperte, "a few hundred agents" attivi durante il giorno, "a few thousand" di notte per lavori più approfonditi.
38. Sub-agents. + slash command /loop = nuova primitiva principale.
39. *"Loops are the future at this point. If you haven't experimented with it, highly highly recommend it.". * ### Loop: la nuova primitiva
40. Meccanismo. "You have Claude use cron to schedule a job for some point in the future, and it's a repeat job."
41. Cadenze. ogni minuto / ogni 5 minuti / ogni giorno — configurabile.
42. Esempi di Cherny.
43. Sorvegliante di PR. "fixing CI, auto-rebasing"
44. Custode della salute della CI. "if there's a flaky test or whatever, it'll go and fix it"
45. Raggruppatore di feedback Twitter. "grabs feedback from Twitter and clusters it for me every 30 minutes"
46. Routines. = equivalente lato server del loop: "we just launched routines. Even if you close your laptop, it keeps going."
47. 4.7 avvia i loop spontaneamente. "I'll tell it 'pull this data query.' And it's like 'Hey, I noticed the data is changing over time. I'll start a loop and I'll give you a report every 30 minutes.' And I'm like 'Great. Can you send it to me over Slack?' And then it uses the Slack MCP to do that." ### Visione sui team
48. "There's going to be a lot more generalists than there are today."
49. Non solo "product engineers cross-platform" (iOS+web+server) ma generalisti interdisciplinari: "engineers that are really good at product engineering, but also really great at design. Or really great at product and data science and engineering."
50. Nel team di Claude Code.
51. "Every single person on our team writes code."
52. Elenco: engineering manager, PM, designer, data scientist, addetto alla finanza, user researcher.
53. "They're specialist in something, but now also everyone's just coding." ### Visione sul SaaS / disruption
54. Nessuna apocalisse del SaaS. , ma un riordino tramite il framework delle 7 Powers di Hamilton Helmer.
55. Powers che diventano meno importanti.
56. Switching costs. — "because you can just use the model and you can port from one thing to a different thing."
57. Process power. — "Claude is getting really good at figuring out process. And especially with 4.7, it can just hill climb anything. So if you give it a target and you tell it to iterate until it's done, it will just do it. I think this is the first model like that."
58. Powers che restano importanti.
59. Network effects.
60. Scale economies.
61. Cornered resources.
62. Previsione sulle startup. "the number of startups in the next 10 years that are just going to disrupt everything is going to increase like 10x."
63. Perché ora. "a tiny startup can build a thing that's as valuable as a large company and actually compete head-to-head, because the large company has to evolve their business process, retrain everyone, face internal resistance. But you, starting fresh, can build with AI natively from the ground up."
64. "It's the best time to build. It's the best time to be a startup. There's so much disruption coming." ### Analogia del pivot — la stampa di Gutenberg
65. L'Europa del 1400 prima della stampa. "essentially 10% of the European population was literate. They knew how to read and write. They were often employed by kings and lords that were not literate."
66. 50 anni dopo l'invenzione. "more literature published in Europe than in the thousand years before."
67. Costo di un libro. ↓ 100 volte nello stesso periodo.
68. Qualche secolo dopo. "literacy globally went up to like 70%. Now we can all read and write. You don't need a degree in reading and writing to know how to read and write. Although still there are professional writers."
69. Analogia applicata al 2026+. "software will be a thing that is fully democratized, that anyone can do. And it's going to be much faster than 50 years."
70. Frase pivot memorabile. "The best person to write accounting software, I think maybe even today, is not an engineer, it's a really good accountant. Because they know the domain really well, and coding is the easy part. It's knowing the domain that's the hard part." ### Modello vs harness (Q&A)
71. "If you asked me a year ago, the ratio was maybe 50/50. If you asked me 6 months ago, would be 50/50. In 2 years? I don't know dude, we plan in like 1 week out."
72. L'harness diventa meno importante. man mano che il modello migliora.
73. Evoluzioni dell'harness.
74. Loop come cittadini di prima classe
75. Più facile eseguire molti agenti
76. Sub-agents (un'idea tra le altre)
77. "There's a bunch more stuff that we're cooking."
78. Tra 1 anno. "the model will be much better aligned. And so all the safety mechanisms that we have today around prompt injection and static verification of commands and permission modes, human in the loop, all this kind of stuff is just going to be less important because the model will just do the right thing." ### Democratizzazione: tutti programmano come tutti sanno mandare un SMS
79. "I think it's going to be a skill like — yeah, like I know how to send a text message."
80. Forte richiamo al modello della stampa di Gutenberg citato sopra. ### Pratiche interne di Anthropic
81. Nessun vantaggio di modello. "internally, we use the same models everyone else does." Mythos per i test, molto Opus 4.7 per il dogfooding.
82. Pratica radicale.
83. "We use Claude for literally everything."
84. "Our Claudes are talking all day. As I'm coding, as my Claudes are coding in a loop, they will communicate over Slack to talk to other people's Claudes that are also running in a loop, to figure out unknowns."
85. "We have no more manually written code anywhere at the company."
86. "All of the SQL is written by models. Everything is just built by the models."
87. Il vantaggio di Anthropic non è la tecnologia ma il processo. "the place that we're ahead is not the technology — same technology available to us is available to everyone here, because fundamentally we are building a platform. There's actually a far bigger lead in organizational structure and organizational process." ### Multi-agent / parallelizzazione
88. "On the product side, it really just comes down to prompting. As the model gets better, it just naturally does this."
89. "It's not on users to figure out how to hold the tools better. If that's the case, it's actually a product design problem and I'm not doing a good job." ### Locale vs cloud, computer use, MCP
90. Locale vs cloud: "doesn't matter. By a couple years from now, the model is just going to be doing all the code. It's going to be starting the agents. It's going to be building the environments. If it decides 'actually I'll use local models to do this,' that's what it'll do. These I don't think these will be decisions that we are making as engineers anymore."
91. Knowledge work: "largely, it's there already with Salesforce and Docs. For us, it's always just the simplest answer. It's MCP."
92. Computer use: "I think currently, Anthropic is pretty far ahead on computer use. It's quite good. It's very slow, but it does it quite well now, especially with 4.7."
93. "All this stuff just doesn't matter that much. It could be MCPs, APIs, just some sort of programmatic access — to the model, it's just tokens." ### Prodotti in arrivo
94. Claude Design. "It's pretty good today. It's going to get a lot better."
95. Claude Code. "There's also a few things that we're cooking up for Claude Code that are going to be landing over the coming weeks."
96. Loop, batch, parallelismo massiccio di agenti. "that's going to get better."
97. Computer use. . ### Collegamento con la cartella di veille
98. Quarta fiche su Cherny. nella cartella, dopo:
99. cherny-yc-lightcone-claude-code-origin-story-2026-02. (genesi YC Lightcone)
100. cherny-claude-code-10-tips-team-x-2026-02-01. (10 consigli su X)
101. nunez-cherny-claude-code-workflow-venturebeat-2026-01-05. (workflow VentureBeat)
102. wu-cherny-use-claude-code-builders-every-2025-10-29. (Wu/Cherny Every)
103. Inoltre la fiche anthropic-claude-code-quality-postmortem-2026-04-23, di cui Cherny è uno degli autori impliciti.
104. Contributo specifico di questa fiche. rispetto alle precedenti:
105. Configurazione solo iOS. = novità (rispetto alla configurazione CLI/desktop delle fiche precedenti).
106. /loop + Routines. = nuova primitiva di architettura agentica presentata pubblicamente.
107. Il racconto "incubatore Anthropic Labs + round 2 di Mike Krieger". = nuovo dettaglio organizzativo.
108. Il punto di svolta di Opus 4, maggio 2025. = datazione precisa del decollo.
109. "100% di codice generato da ottobre/novembre 2025". = datazione personale.
110. L'analogia di Gutenberg. + "il miglior commercialista scrive il miglior software di contabilità" = frasi pivot memorabili.
111. Il riordino delle "7 Powers" di Helmer. = analisi strutturata del SaaS.
112. Sequoia / Lauren Reader. = nuova sede, pubblico di investitori.
113. Convergenza principale.
114. Gli agenti cron /loop. = implementazione radicale del Ralph Loop (Trivedy 2026-03-10), del Compound Engineering di Every (Shipper/Klaassen 2025-12-11), e dei Compounding Teams (Schillace 2025-09-28).
115. "Claude che parlano con Claude su Slack". = implementazione pratica delle previsioni di Levie in Building for Trillions of Agents (2026-03-07) e dell'infrastruttura agent-native sostenuta da Karpathy (2026-04-29).
116. "Coding is solved" + 100% di codice generato + 150 PR/giorno. si allinea con Curran/Intercom, produttività R&D 3x (2026-04-16, 93,6% di PR guidate da agenti), Stripe Minions (2026-02-09 e 2026-02-19), StrongDM Software Factory (2026-02-06).
117. La democratizzazione in stile Gutenberg. = stessa tesi di Karpathy (2026-04-29) sui neural computers e di Andreessen sulla ridefinizione del programmatore (2026-02). Coerente anche con il progetto "tutti possono programmare" sostenuto da Wescale (2026-05-03) e Habert PROJ-AI (2026-05-05).
118. Le "7 Powers" e il riordino del SaaS. = espongono un framework di strategia aziendale raro nel corpus 2026, complementare a Permanent Underclass di Sun NYT (2026-04-30) sul versante economico e alla Tension Map di Pawlowski (2026-03-30) sul versante della ricerca di mercato.
119. "L'harness diventa meno importante man mano che il modello migliora". = una sfumatura diretta rispetto alla tendenza Harness Engineering (Trivedy 2026-03-10, Osmani 2026-04-19, Böckeler 2026-04-02). Cherny anticipa il progressivo assorbimento dell'harness nel modello allineato.
120. "Il miglior commercialista scrive il miglior software di contabilità". = un corollario della long-tail dei bisogni aziendali anticipata dalle fiche Wescale (2026-05-03, "long-standing needs that were too costly can finally be addressed") e dal Developer Taste di Soto (2026-04).
121. Limiti da segnalare.
122. Ottimismo tecnocentrico deliberato. — Cherny è il creatore di Claude Code, un evidente bias da evangelista sul "coding is solved."
123. Nessuna cifra esterna. — si basa sulla propria pratica e su quella di Anthropic. Da confrontare con le cifre rigorose di Curran/Intercom (R&D 3x, 93,6% di PR guidate da agenti).
124. Formato Sequoia. nessuna domanda contraddittoria, atmosfera amichevole, nessuna messa in discussione della jagged intelligence (Karpathy 2026-04-29) o del permanent underclass (Sun NYT 2026-04-30).
125. "Coding is solved". un'affermazione provocatoria che merita di essere sfumata dalla frase di Karpathy "Opus 4.7 will simultaneously refactor a 100,000 line codebase and tell me to walk to a car wash 50 meters away" (2026-04-29) — il coding di superficie è effettivamente automatizzabile, ma il giudizio sistemico resta un dominio difficile.
126. Utile per. una presentazione retrospettiva di 18 mesi su Claude Code (narrazione + datazioni); un argomento "best time to build" per una startup davanti a un comitato esecutivo; una presentazione "i loop come futuro" per i team di ingegneria; un corso/talk sulla "democratizzazione del software tramite l'analogia di Gutenberg"; un dibattito sulla posizione dell'harness nella dottrina 2026.