Historically, Sonnet has been Opus’s cheaper, faster sibling: You traded some brainpower for speed and cost savings.

Now, with the release of Sonnet 4.6, Anthropic says you don’t have to trade anything—just pay less. If you’re running a live app with users on Opus right now, there is some good news about your API costs: Sonnet 4.6 costs $3 input/$15 output per million tokens—roughly half what Opus runs. (GPT-5.2, for comparison, costs $1.75 input/$14 output per 1 million tokens.) Cost has been a consistent sticking point with Opus, and with Sonnet 4.6, Anthropic seems to be saying, “We hear you.”

These are true day-zero tests—we got access when everyone else did—so let’s dive in to determine in real time whether it’s true that Sonnet’s new capabilities come without sacrifice. You can also watch the livestream we did yesterday as soon as the model dropped.

The Reach Test

Dan Shipper, cofounder and CEO 🟨

“Green for using inside of products you’re building, yellow for my own daily work. If you’re building an app that was previously too expensive to run on Opus, this is a big deal. For coding and complex tasks where I can afford Opus and 5.3 Codex, I’d probably stick with it.”

Kieran Klaassen, general manager of Cora 🟨

“It’s very usable, and for a lot of people it should be the default. But it frustrated me when it got stuck on something Opus solved in one shot. Anytime a model frustrates me, that’s a no-go for my personal workflow—but this is genuinely solid for production.”

What works well

The model is smart. Across coding tasks, pull request triage, brainstorming, and a complex P&L restructuring, it held the thread, followed through on multi-step instructions, and didn’t make the kind of mid-task mistakes that would have tripped up Sonnet 4.5.

Kieran ran it through a full compound engineering workflow—sorting through a backlog of pending code changes, merging branches, writing changelogs, and theming issues—and came away impressed. “I’ve not found any reasons to believe it’s not as smart as Opus 4.6,” he said.

The economics are a big win for teams using Opus inside production AI apps. Spiral, Every’s AI ghostwriter, for example, has been running on Opus at up to $1,000 a day in token costs. With Sonnet 4.6, that cost roughly halves, without touching the codebase.

What needs work

Given past precedent, you’d expect a new Sonnet model to be meaningfully faster than Opus. This one feels nearly the same. That’s fine if you’re getting Opus-quality output, but disappointing if you were counting on snappier performance for iteration-heavy workflows.

The model also showed some erratic behavior under pressure. When Dan asked it to plan a homepage redesign, it correctly asked for a name for the work tree—a separate copy of the codebase where it could experiment safely—then immediately started rewriting the homepage anyway. As Dan noted, “It’s both too cautious and too eager.” And when Kieran ran into an MCP configuration issue, Sonnet 4.6 spun in circles while Opus solved it in one shot and cited the exact wrong file.

The verdict

Dan thinks that Anthropic’s strategy is to keep model tier prices stable while continuously improving the product. Whatever Opus can do today, Sonnet will be able to do it in six months for less.

If you’re building AI-powered products and have been holding off on Opus because of cost, the wait is over. Sonnet 4.6 appears to be the model that delivers the capability you want without making your bank account cry. If you’re using Claude for your own complex work and can swing the Opus price, you probably don’t need to switch yet. But if you’re on the $20 Pro plan and want more horsepower, this is a meaningful upgrade.

For more on Claude Sonnet, check out our previous coverage:

***Katie Parrott*** *is a staff writer and AI editorial lead at Every. You can read more of her work in* *her newsletter.*\n

To read more essays like this, subscribe to Every, and follow us on X at @every and on LinkedIn.

We build AI tools for readers like you. Write brilliantly with Spiral. Organize files automatically with Sparkle. Deliver yourself from email with Cora. Dictate effortlessly with Monologue.

We also do AI training, adoption, and innovation for companies. Work with us to bring AI into your organization.

Get paid for sharing Every with your friends. Join our referral program.

For sponsorship opportunities, reach out to sponsorships@every.to.

Help us scale the only subscription you need to stay at the edge of AI. Explore open roles at Every.

","freeContentEncoded","textContent","published","strategy","curl-ready","locked","deletedAt","isSyncedFromOwnBlog","original","duplicatesCount","duplicateVersion","originalArticleId","errorsCount","appearance","article","color","backgroundColor","shortSummary","Is the new Sonnet 4.6 really as good as Opus, but cheaper? We put it to the test!","summary","Anthropic's Sonnet 4.6 model offers performance comparable to its more expensive sibling, Opus, at roughly half the price, making it a cost-effective option for live applications. While it excels in complex tasks and multi-step instructions, it doesn't offer the expected speed improvements and can exhibit erratic behavior under pressure. For new AI product development, Sonnet 4.6 is presented as a compelling choice, while existing Opus users for complex personal work may not need to switch immediately.","bullets",[166,167,168,169,170,171],"Sonnet 4.6 is priced at $3 input/$15 output per million tokens, significantly cheaper than Opus.","The model performs well on coding, brainstorming, and complex restructuring tasks, maintaining a thread and following instructions.","While not significantly faster than Opus, Sonnet 4.6 offers Opus-quality output at a lower cost, making it suitable for production AI apps.","The model showed some erratic behavior, such as starting a task before confirmation or getting stuck on issues Opus solved quickly.","Anthropic's strategy may involve continuously improving Sonnet to match Opus capabilities over time at a stable, lower price point.","Sonnet 4.6 is recommended for developers building AI products concerned about cost and for users on the $20 Pro plan seeking more power.","persons",[174,175],"Kieran","Dan","places",[],"words",653,"sentiment",1.5,"unknown","subject","tech",{"_186":187,"_188":189,"_216":197},"type","bee:get-index-of-articles","schedule",{"_19":190,"_37":-5,"_191":192,"_63":-5,"_193":-5,"_194":195,"_196":197,"_198":56,"_66":67,"_183":-5,"_199":-5,"_200":67,"_146":147,"_201":202,"_39":202,"_203":204,"_205":56,"_206":207,"_41":202,"_208":67,"_65":-5,"_209":210,"_211":-5,"_212":213,"_214":-5,"_215":-5},"2db9aa18-5531-49c9-b5f5-d910a3f31f21","npub","npub1uxmvt46kwqzen4vsnd6zle7km3rg34k067tgshlcyhlpz0t58zlq6gaspr","jobId","order",1,"rssUrl","https://every.to/","system","userRef","autosign","syncedAt","2026-04-02T13:32:24.297Z","frequency","daily","isOwnBlog","nextRunAt","2026-04-02T13:32:35.660Z","autopublish","sourceVariant","rss","systemVariant","scheduleVariant","scraping-bee-classic-proxy","retryArticlesIds","ownershipConfirmedAt","dataSourceUrl","eventId","9ef47e095133a7d1f8cc3172184ef65a6ab5825a96c6388b427d6028131f7559","nostrRootEvent","event",{"_19":218,"_222":223,"_224":195,"_63":225,"_237":236,"_139":238,"_239":240},"sig","9bbd7aa354e4dc48b6845933d267694d79f203af1e592143ef9b970d30ffde5b7f0340188a772a0c5d5f4ee8ba7eb5779ec3dfb349d9d0bc1cf6a6d312dca7bc","kind",[226,227,228,232,234],[183,125],[32,132],[229,230,16,231],"a","30023:e1b6c5d756700599d5909b742fe7d6dc4688d6cfd796885ff825fe113d7438be:088db8fc-5d6f-41f9-aaaf-f9a2dc6cfbbc","mention",[233,16],"client",[235,236,16],"p","e1b6c5d756700599d5909b742fe7d6dc4688d6cfd796885ff825fe113d7438be","pubkey","Vibe Check: Anthropic Just Made Opus Cheaper Without Calling It That\nSonnet 4.6 delivers Opus-close performance at half the price—but speed didn't come along for the ride\n\n\nnostr:naddr1qvzqqqr4gupzpcdkcht4vuq9n82epxm59lnadhzx3rtvl4uk3p0lsf07zy7hgw97qqjrqwpcv33rsenr956kgdnx956rze3e94skzctx94nrjcfjv33nvcmxvf3xxkwanyf ","created_at",1775136798,"authorNPub","markdown","Historically, Sonnet has been Opus’s cheaper, faster sibling: You traded some brainpower for speed and cost savings.\n\nNow, with the release of Sonnet 4.6, Anthropic says you don’t have to trade anything—just pay less. If you’re running a live app with users on Opus right now, there is some good news about your API costs: Sonnet 4.6 costs $3 input/$15 output per million tokens—roughly half what Opus runs. (GPT-5.2, for comparison, costs $1.75 input/$14 output per 1 million tokens.) Cost has been a consistent sticking point with Opus, and with Sonnet 4.6, Anthropic seems to be saying, “We hear you.”\n\nThese are true day-zero tests—we got access when everyone else did—so let’s dive in to determine in real time whether it’s true that Sonnet’s new capabilities come without sacrifice. You can also [watch the livestream](https://www.youtube.com/watch?v=mahp6-THjuY) we did yesterday as soon as the model dropped.\n\nThe Reach Test\n--------------\n\n**Dan Shipper, cofounder and CEO 🟨**\n\n“Green for using inside of products you’re building, yellow for my own daily work. If you’re building an app that was previously too expensive to run on Opus, this is a big deal. For coding and complex tasks where I can afford Opus and 5.3 Codex, I’d probably stick with it.”\n\n**Kieran Klaassen, general manager of [Cora](https://cora.computer) 🟨**\n\n“It’s very usable, and for a lot of people it should be the default. But it frustrated me when it got stuck on something Opus solved in one shot. Anytime a model frustrates me, that’s a no-go for my personal workflow—but this is genuinely solid for production.”\n\n### What works well\n\nThe model is smart. Across coding tasks, pull request triage, brainstorming, and a complex P&L restructuring, it held the thread, followed through on multi-step instructions, and didn’t make the kind of mid-task mistakes that would have tripped up Sonnet 4.5.\n\nKieran ran it through a full [compound engineering](https://every.to/chain-of-thought/compound-engineering-how-every-codes-with-agents) workflow—sorting through a backlog of pending code changes, merging branches, writing changelogs, and theming issues—and came away impressed. “I’ve not found any reasons to believe it’s not as smart as [Opus 4.6](https://every.to/vibe-check/opus-4-6),” he said.\n\nThe economics are a big win for teams using Opus inside production AI apps. [**Spiral**](https://writewithspiral.com), Every’s AI ghostwriter, for example, has been running on Opus at up to $1,000 a day in token costs. With Sonnet 4.6, that cost roughly halves, without touching the codebase.\n\n### What needs work\n\nGiven past precedent, you’d expect a new Sonnet model to be meaningfully faster than Opus. This one feels nearly the same. That’s fine if you’re getting Opus-quality output, but disappointing if you were counting on snappier performance for iteration-heavy workflows.\n\nThe model also showed some erratic behavior under pressure. When Dan asked it to plan a homepage redesign, it correctly asked for a name for the work tree—a separate copy of the codebase where it could experiment safely—then immediately started rewriting the homepage anyway. As Dan noted, “It’s both too cautious and too eager.” And when Kieran ran into an MCP configuration issue, Sonnet 4.6 spun in circles while Opus solved it in one shot and cited the exact wrong file.\n\n### The verdict\n\nDan thinks that Anthropic’s strategy is to keep model tier prices stable while continuously improving the product. Whatever Opus can do today, Sonnet will be able to do it in six months for less.\n\nIf you’re building AI-powered products and have been holding off on Opus because of cost, the wait is over. Sonnet 4.6 appears to be the model that delivers the capability you want without making your bank account cry. If you’re using Claude for your own complex work and can swing the Opus price, you probably don’t need to switch yet. But if you’re on the $20 Pro plan and want more horsepower, this is a meaningful upgrade.\n\n* * *\n\n_For more on Claude Sonnet, check out our previous coverage:_\n\n1. [“Vibe Check: Claude 3.7 Sonnet and Claude Code](https://every.to/vibe-check/vibe-check-claude-3-7-sonnet-and-claude-code)”\n2. [“Vibe Check: Sonnet 4.5](https://every.to/vibe-check/vibe-check-claude-sonnet-4-5)”\n3. [“We Tested Claude Sonnet 4.5 for Writing and Editing](https://every.to/vibe-check/vibe-check-we-tested-claude-sonnet-4-5-for-writing-and-editing)”\n4. [“Claude Sonnet 4 Now Has a 1-million Token Context Window](https://every.to/vibe-check/vibe-check-claude-sonnet-4-now-has-a-1-million-token-context-window)”\n\n* * *\n\n\\*\\*\\*[Katie Parrott](https://every.to/@katie.parrott12)\\*\\*\\* \\*is a staff writer and AI editorial lead at Every. You can read more of her work in\\* \\*[her newsletter](https://katieparrott.substack.com/).\\*\n\n_To read more essays like this, subscribe to [Every](https://every.to/subscribe), and follow us on X at [@every](http://twitter.com/every) and on [LinkedIn](https://www.linkedin.com/company/everyinc/)._\n\n_We [build AI tools](https://every.to/studio) for readers like you. Write brilliantly with [Spiral](https://writewithspiral.com/). Organize files automatically with **[Sparkle](https://makeitsparkle.co/?utm_source=everyfooter)**. Deliver yourself from email with [Cora](https://cora.computer/). Dictate effortlessly with **[Monologue](https://monologue.to/)**._\n\n_We also do AI training, adoption, and innovation for companies. [Work with us](https://every.to/consulting?utm_source=emailfooter) to bring AI into your organization._\n\n_Get paid for sharing Every with your friends. Join our [referral program](https://every.getrewardful.com/signup)._\n\n_For sponsorship opportunities, reach out to sponsorships@every.to._\n\n_Help us scale the only subscription you need to stay at the edge of AI. Explore [open roles at Every](https://www.notion.so/Jobs-Every-25cca4f355ac80c5ad6ee7a6e93d6b4e?pvs=21)._","freeMarkdown","kind1EventId","kind30023EventId","f2a74a3e969f1b45cbd5c6b1211e776a3077420ee8c8ba3558aa045e404f3322","kind30023",{"_19":247,"_222":250,"_224":251,"_63":252,"_237":236,"_139":266,"_239":267},"b8063b31264990b63e8c40169edd34a4d044956fdcd1f8bc2a594732cb7121f770edc8f66bbabd02c1da48def6e30d6f12a2c27c4e4f151dae9dba6fa5664ba0",30023,[253,256,257,260,261,263,264,265],[254,255],"r","https://meta.layer3.press/articles/088db8fc-5d6f-41f9-aaaf-f9a2dc6cfbbc",[30,125],[258,259],"published_at","1775136837",[32,132],[262,120],"d",[162,126],[233,16],[235,236,16],"Vibe Check: Anthropic Just Made Opus Cheaper Without Calling It That\nAnthropic's Sonnet 4.6 model offers performance comparable to its more expensive sibling, Opus, at roughly half the price, making it a cost-effective option for live applications. While it excels in complex tasks and multi-step instructions, it doesn't offer the expected speed improvements and can exhibit erratic behavior under pressure. For new AI product development, Sonnet 4.6 is presented as a compelling choice, while existing Opus users for complex personal work may not need to switch immediately.\n - Sonnet 4.6 is priced at $3 input/$15 output per million tokens, significantly cheaper than Opus.\n - The model performs well on coding, brainstorming, and complex restructuring tasks, maintaining a thread and following instructions.\n - While not significantly faster than Opus, Sonnet 4.6 offers Opus-quality output at a lower cost, making it suitable for production AI apps.\n - The model showed some erratic behavior, such as starting a task before confirmation or getting stuck on issues Opus solved quickly.\n - Anthropic's strategy may involve continuously improving Sonnet to match Opus capabilities over time at a stable, lower price point.\n - Sonnet 4.6 is recommended for developers building AI products concerned about cost and for users on the $20 Pro plan seeking more power.\n",1775136837,"publishedToNostr","htmlContentKind30023","

Historically, Sonnet has been Opus’s cheaper, faster sibling: You traded some brainpower for speed and cost savings.

The Reach Test

Dan Shipper, cofounder and CEO 🟨

Kieran Klaassen, general manager of Cora 🟨

What works well

What needs work

The verdict

Dan thinks that Anthropic’s strategy is to keep model tier prices stable while continuously improving the product. Whatever Opus can do today, Sonnet will be able to do it in six months for less.

For more on Claude Sonnet, check out our previous coverage:

***Katie Parrott*** *is a staff writer and AI editorial lead at Every. You can read more of her work in* *her newsletter.*

To read more essays like this, subscribe to Every, and follow us on X at @every and on LinkedIn.

We build AI tools for readers like you. Write brilliantly with Spiral. Organize files automatically with Sparkle. Deliver yourself from email with Cora. Dictate effortlessly with Monologue.

We also do AI training, adoption, and innovation for companies. Work with us to bring AI into your organization.

Get paid for sharing Every with your friends. Join our referral program.

For sponsorship opportunities, reach out to sponsorships@every.to.

Help us scale the only subscription you need to stay at the edge of AI. Explore open roles at Every.

","pubDate","2026-04-02T13:33:57.864Z","modifiedAt","2026-03-30T00:00:00.000Z","2026-04-02T13:33:18.064Z","2026-04-02T13:33:57.663Z","storyId","likes","zaps","reposts","comments","storyUpdatedAt","storyArticlesCount",{"_19":285,"_121":286,"_123":286,"_124":13,"_30":287,"_31":288,"_127":287,"_128":288,"_129":-5,"_130":-5,"_131":289,"_133":-5,"_134":-5,"_135":-5,"_136":-5,"_137":-5,"_138":-5,"_139":-5,"_141":-5,"_143":-5,"_144":-5,"_21":145,"_146":147,"_148":-5,"_69":56,"_149":-5,"_150":56,"_151":-5,"_152":94,"_153":-5,"_154":-5,"_155":94,"_156":157,"_158":-5,"_159":-5,"_160":290,"_162":291,"_164":292,"_172":296,"_176":297,"_178":298,"_180":94,"_65":182,"_183":184,"_24":299,"_63":-5,"_217":301,"_219":-5,"_220":302,"_241":-5,"_242":-5,"_244":-5,"_245":301,"_246":312,"_248":313,"_268":56,"_269":-5,"_271":328,"_273":329,"_39":275,"_41":330,"_42":-5,"_277":-5,"_278":-5,"_279":-5,"_280":-5,"_281":94,"_282":-5,"_283":-5},"08756d26-b3f4-45f4-a0b5-98ec9d957e3b","https://every.to/source-code/you-have-a-claw-now-what","You Have a Claw. Now What?","Our comprehensive guide to OpenClaw is here","https://d24ovhgu8s7341.cloudfront.net/uploads/post/social_media_image/3958/full_page_cover_OpenClawOur_Comprehensive_Guide_for_Beginners(2).png","Everything you need to know about OpenClaw in one place!","A detailed guide to OpenClaw has been released. It aims to provide comprehensive information on the subject. This resource is designed to help users understand and utilize OpenClaw effectively.",[293,294,295],"A comprehensive guide to OpenClaw is now available.","This guide serves as a complete resource for understanding OpenClaw.","It is designed to assist users with OpenClaw.",[],[],21,{"_186":187,"_188":300,"_216":197},{"_19":190,"_37":-5,"_191":192,"_63":-5,"_193":-5,"_194":195,"_196":197,"_198":56,"_66":67,"_183":-5,"_199":-5,"_200":67,"_146":147,"_201":202,"_39":202,"_203":204,"_205":56,"_206":207,"_41":202,"_208":67,"_65":-5,"_209":210,"_211":-5,"_212":213,"_214":-5,"_215":-5},"58227e0f30d2715699ff658892eef7cf9a6e26c5d086192b722ca1aeb255e177",{"_19":301,"_222":303,"_224":195,"_63":304,"_237":236,"_139":311,"_239":240},"adad9d65b89761337a219b076ab67705cd3baf4dc095ccefde63491826721ebe1b46d62062b3ab9b9f37aba5530ca252c9bfea7682c9581bca34f2f49e58a120",[305,306,307,309,310],[183,287],[32,289],[229,308,16,231],"30023:e1b6c5d756700599d5909b742fe7d6dc4688d6cfd796885ff825fe113d7438be:08756d26-b3f4-45f4-a0b5-98ec9d957e3b",[233,16],[235,236,16],"You Have a Claw. Now What?\nOur comprehensive guide to OpenClaw is here\n\n\nnostr:naddr1qvzqqqr4gupzpcdkcht4vuq9n82epxm59lnadhzx3rtvl4uk3p0lsf07zy7hgw97qqjrqwphx5mxgv3k943rxe35956r2e3594snqc3495unsetr89jrjdfhv5ekyhdsrj6 ","72142d3e0c9971f0a33db522a6d1af7a0ef26367f4708c58078c0df1f2bbdc9b",{"_19":312,"_222":314,"_224":251,"_63":315,"_237":236,"_139":326,"_239":327},"2518303528f374fddfbc27b25a3ad4666ffdd7a474ffdc96e8f82b11d4dcac3432fb0e1261084f1e4581af90eb89c389b675275719da97e7e812116cca7d940b",[316,318,319,321,322,323,324,325],[254,317],"https://meta.layer3.press/articles/08756d26-b3f4-45f4-a0b5-98ec9d957e3b",[30,287],[258,320],"1775136836",[32,289],[262,285],[162,288],[233,16],[235,236,16],"You Have a Claw. Now What?\nA detailed guide to OpenClaw has been released. It aims to provide comprehensive information on the subject. This resource is designed to help users understand and utilize OpenClaw effectively.\n - A comprehensive guide to OpenClaw is now available.\n - This guide serves as a complete resource for understanding OpenClaw.\n - It is designed to assist users with OpenClaw.\n",1775136836,"2026-04-02T13:33:56.607Z","2026-03-24T00:00:00.000Z","2026-04-02T13:33:56.325Z",{"_19":332,"_121":333,"_123":333,"_124":13,"_30":334,"_31":335,"_127":334,"_128":335,"_129":-5,"_130":-5,"_131":336,"_133":-5,"_134":-5,"_135":-5,"_136":-5,"_137":-5,"_138":-5,"_139":337,"_141":338,"_143":-5,"_144":-5,"_21":145,"_146":147,"_148":-5,"_69":56,"_149":-5,"_150":56,"_151":-5,"_152":94,"_153":-5,"_154":-5,"_155":94,"_156":157,"_158":-5,"_159":-5,"_160":339,"_162":340,"_164":341,"_172":347,"_176":350,"_178":351,"_180":352,"_65":182,"_183":184,"_24":353,"_63":-5,"_217":355,"_219":-5,"_220":356,"_241":-5,"_242":366,"_244":-5,"_245":355,"_246":367,"_248":368,"_268":56,"_269":383,"_271":384,"_273":385,"_39":275,"_41":386,"_42":-5,"_277":-5,"_278":-5,"_279":-5,"_280":-5,"_281":94,"_282":-5,"_283":-5},"0265b3b6-e584-4b07-a74a-2516fb4edc7f","https://every.to/context-window/everyone-gets-a-sidekick","Everyone Gets a Sidekick","Plus: The best from our Q2 Demo Day","https://d24ovhgu8s7341.cloudfront.net/uploads/post/social_media_image/4087/full_page_cover_CW_Sunday_Cover_Image(2).png","I stare at my screen some days and think: Why hasn’t AI replaced me yet?\n\nI spend my hours playing textual Tetris—nudging workstreams, reviewing code, editing prose, shipping features. An AI agent can do all of those things. So why am I still here?\n\nBecause taste can’t be typed out. It has to be worn in.\n\nIf I said to our managing editor, Eleanor Warnock, “Write down everything I’d need to edit one of our pieces,” it would be impossible. Her instincts come from hundreds of past edits, thousands of small decisions layered on top of each other. You would need to work with Eleanor for a long time to emulate her editing style. You can’t enumerate it from scratch.\n\nThe gap between what I want and what AI gives me is real. To get a result I’m satisfied with, I need what I’ve always needed: time. I lean on AI to make a decision, but it’s not the decision I would make. So I give feedback. Then I give it again. And again. The process of teaching an AI your taste looks a lot like the process of developing taste in the first place—the accumulation of many small moments, each one building like sediment on the last.\n\nEvery’s AI-native engineering philosophy, compound engineering, recognizes this need for ongoing growth. After every piece of work, you ask your AI to distill and integrate the lessons you’ve learned. The next time you encounter a similar problem, you’re better able to solve it. After many cycles, you amass a war chest of small opinions. The AI may be fast, but there’s no way to speedrun the process.\n\nThe same goes for trust. People start timidly with OpenClaw, asking it to do simple tasks. Then, they give the agent a little more responsibility. When it does well, they share a bit more context, grant a bit more permission. The output improves. Trust builds, the same way it builds with any human: one kept promise at a time.\n\nThat’s good reason to start now. A person who spends time with their AI today, accreting those layers of context and taste and trust, will be meaningfully ahead of someone who starts next year.\n\nAI can help us move faster between the moments that matter. But it can’t manufacture the moments themselves. Some things won’t be rushed. I remain uncompressed.—*Willie Williams*","

I stare at my screen some days and think: Why hasn’t AI replaced me yet?

\n\n

I spend my hours playing textual Tetris—nudging workstreams, reviewing code, editing prose, shipping features. An AI agent can do all of those things. So why am I still here?

\n\n

Because taste can’t be typed out. It has to be worn in.

\n\n

If I said to our managing editor, Eleanor Warnock, “Write down everything I’d need to edit one of our pieces,” it would be impossible. Her instincts come from hundreds of past edits, thousands of small decisions layered on top of each other. You would need to work with Eleanor for a long time to emulate her editing style. You can’t enumerate it from scratch.

\n\n

The gap between what I want and what AI gives me is real. To get a result I’m satisfied with, I need what I’ve always needed: time. I lean on AI to make a decision, but it’s not the decision I would make. So I give feedback. Then I give it again. And again. The process of teaching an AI your taste looks a lot like the process of developing taste in the first place—the accumulation of many small moments, each one building like sediment on the last.

\n\n

Every’s AI-native engineering philosophy, compound engineering, recognizes this need for ongoing growth. After every piece of work, you ask your AI to distill and integrate the lessons you’ve learned. The next time you encounter a similar problem, you’re better able to solve it. After many cycles, you amass a war chest of small opinions. The AI may be fast, but there’s no way to speedrun the process.

\n\n

The same goes for trust. People start timidly with OpenClaw, asking it to do simple tasks. Then, they give the agent a little more responsibility. When it does well, they share a bit more context, grant a bit more permission. The output improves. Trust builds, the same way it builds with any human: one kept promise at a time.

\n\n

That’s good reason to start now. A person who spends time with their AI today, accreting those layers of context and taste and trust, will be meaningfully ahead of someone who starts next year.

\n\n

AI can help us move faster between the moments that matter. But it can’t manufacture the moments themselves. Some things won’t be rushed. I remain uncompressed.—Willie Williams

","AI can't replicate human taste or judgment, which are built over time through experience and feedback. Building trust with AI also takes time and consistent positive interactions.","Human intuition and taste, developed over time through numerous small decisions and experiences, cannot be easily replicated by AI. Building trust with AI is similar to building trust with humans, requiring consistent positive interactions and gradual increases in responsibility. Those who invest time in developing their relationship with AI now will be significantly ahead of those who wait.",[342,343,344,345,346],"AI can perform many tasks previously done by humans, but it lacks \"taste\" which is built through experience.","Developing an AI's understanding of taste requires continuous feedback and time, similar to how humans develop taste.","AI-native engineering philosophies emphasize continuous growth and integration of learned lessons.","Trust in AI is built gradually through consistent positive outcomes, much like human trust.","Individuals who start working with AI now will gain a significant advantage over those who begin later.",[348,349],"Eleanor Warnock","Willie Williams",[],319,0.5,{"_186":187,"_188":354,"_216":197},{"_19":190,"_37":-5,"_191":192,"_63":-5,"_193":-5,"_194":195,"_196":197,"_198":56,"_66":67,"_183":-5,"_199":-5,"_200":67,"_146":147,"_201":202,"_39":202,"_203":204,"_205":56,"_206":207,"_41":202,"_208":67,"_65":-5,"_209":210,"_211":-5,"_212":213,"_214":-5,"_215":-5},"fa6007b33db5295c0cc1a83e5ae09377b2f9f433de5d28fd47b979a3e7604bde",{"_19":355,"_222":357,"_224":195,"_63":358,"_237":236,"_139":365,"_239":240},"2f1d197b096f9acd69882f49a9e64e1b5d927feb8d014280f9d568bf08958304cc0bd48f852ad1034134ca7ca1d184c2b8a0d14ef44c4f659c117c1306f2d0b5",[359,360,361,363,364],[183,334],[32,336],[229,362,16,231],"30023:e1b6c5d756700599d5909b742fe7d6dc4688d6cfd796885ff825fe113d7438be:0265b3b6-e584-4b07-a74a-2516fb4edc7f",[233,16],[235,236,16],"Everyone Gets a Sidekick\nPlus: The best from our Q2 Demo Day\n\n\nnostr:naddr1qvzqqqr4gupzpcdkcht4vuq9n82epxm59lnadhzx3rtvl4uk3p0lsf07zy7hgw97qqjrqv3kx43rxc3k94jn2wp5956xyvph94snwdrp95er2vfkve3rgetyvvmkv9k5yeq ","I stare at my screen some days and think: Why hasn’t AI replaced me yet?\n\nI spend my hours playing textual Tetris—nudging workstreams, reviewing code, editing prose, shipping features. An AI agent can do all of those things. So why am I still here?\n\nBecause taste can’t be typed out. It has to be worn in.\n\nIf I said to our managing editor, [Eleanor Warnock](https://every.to/@eleanor_b03474_1), “Write down everything I’d need to edit one of our pieces,” it would be impossible. Her instincts come from hundreds of past edits, thousands of small decisions layered on top of each other. You would need to work with Eleanor for a long time to emulate her editing style. You can’t enumerate it from scratch.\n\nThe gap between what I want and what AI gives me is real. To get a result I’m satisfied with, I need what I’ve always needed: time. I lean on AI to make a decision, but it’s not the decision I would make. So I give feedback. Then I give it again. And again. The process of teaching an AI your taste looks a lot like the process of developing taste in the first place—the accumulation of many small moments, each one building like sediment on the last.\n\nEvery’s AI-native engineering philosophy, [compound engineering](https://every.to/source-code/compound-engineering-the-definitive-guide), recognizes this need for ongoing growth. After every piece of work, you ask your AI to distill and integrate the lessons you’ve learned. The next time you encounter a similar problem, you’re better able to solve it. After many cycles, you amass a war chest of small opinions. The AI may be fast, but there’s no way to speedrun the process.\n\nThe same goes for trust. People start timidly with [OpenClaw](https://every.to/guides/claw-school), asking it to do simple tasks. Then, they give the agent a little more responsibility. When it does well, they share a bit more context, grant a bit more permission. The output improves. Trust builds, the same way it builds with any human: one kept promise at a time.\n\nThat’s good reason to start now. A person who spends time with their AI today, accreting those layers of context and taste and trust, will be meaningfully ahead of someone who starts next year.\n\nAI can help us move faster between the moments that matter. But it can’t manufacture the moments themselves. Some things won’t be rushed. I remain uncompressed.—[Willie Williams](https://every.to/@williewilliams)","8a3ad4490bd8f9114abd331bb6889909594ed68cda1092a34bd68b9d9e05f0cf",{"_19":367,"_222":369,"_224":251,"_63":370,"_237":236,"_139":381,"_239":382},"fbee49f53564d260e1bb216c085c4a5015661deb83382e1fd1d2a8c36a23f9e13ce3f009d9d396b0619a95153177203b1c71f3b941be889790ec523c9fc3dd18",[371,373,374,376,377,378,379,380],[254,372],"https://meta.layer3.press/articles/0265b3b6-e584-4b07-a74a-2516fb4edc7f",[30,334],[258,375],"1774742400",[32,336],[262,332],[162,335],[233,16],[235,236,16],"Everyone Gets a Sidekick\nHuman intuition and taste, developed over time through numerous small decisions and experiences, cannot be easily replicated by AI. Building trust with AI is similar to building trust with humans, requiring consistent positive interactions and gradual increases in responsibility. Those who invest time in developing their relationship with AI now will be significantly ahead of those who wait.\n - AI can perform many tasks previously done by humans, but it lacks \"taste\" which is built through experience.\n - Developing an AI's understanding of taste requires continuous feedback and time, similar to how humans develop taste.\n - AI-native engineering philosophies emphasize continuous growth and integration of learned lessons.\n - Trust in AI is built gradually through consistent positive outcomes, much like human trust.\n - Individuals who start working with AI now will gain a significant advantage over those who begin later.\n",1774742400,"

I stare at my screen some days and think: Why hasn’t AI replaced me yet?

I spend my hours playing textual Tetris—nudging workstreams, reviewing code, editing prose, shipping features. An AI agent can do all of those things. So why am I still here?

Because taste can’t be typed out. It has to be worn in.

That’s good reason to start now. A person who spends time with their AI today, accreting those layers of context and taste and trust, will be meaningfully ahead of someone who starts next year.

AI can help us move faster between the moments that matter. But it can’t manufacture the moments themselves. Some things won’t be rushed. I remain uncompressed.—Willie Williams

","2026-03-29T00:00:00.000Z","2026-04-01T00:00:00.000Z","2026-04-02T13:34:05.775Z","nextPage","category","categoryName","Tech","categorySlug","requestUrl","http://data-source-service:3000/public/api/articles?offset=12&limit=13&tags[]=tech&state=published&sortBy=relevance&subject=tech","page",2,"actionData","errors"]