thevibeworks · claude-yolo · May 29, 2026
diff --git a/content/.metadata.json b/content/.metadata.json
diff --git a/content/CHANGELOG.md b/content/CHANGELOG.md
@@ -1,5 +1,9 @@
 # Changelog
 
+## 2.1.156
+
+- Fixed an issue when using Opus 4.8 where thinking blocks were modified, leading to API errors.
+
 ## 2.1.154
 
 - Opus 4.8 is here! Now defaults to high effort · /effort xhigh for your hardest tasks

diff --git a/content/blog/research/impact-software-development.md b/content/blog/research/impact-software-development.md
@@ -7,7 +7,7 @@ Jobs that involve computer programming are a small sector of the modern economy,
 
 In our [previous Economic Index research](https://www.anthropic.com/news/the-anthropic-economic-index), we found very disproportionate use of Claude by US workers in computer-related occupations: that is, there were many more conversations with Claude about computer-related tasks than one would predict from the number of people working in relevant jobs. It’s the same in [the educational context](https://www.anthropic.com/news/anthropic-education-report-how-university-students-use-claude): Computer Science degrees—which involve large amounts of coding—show highly disproportionate AI use.
 
-To understand these changes in more detail, we conducted an analysis of 500,000 coding-related interactions across [Claude.ai](http://claude.ai/redirect/website.v1.fead9975-43b7-403a-972e-eb131ee4485a) (the “default” way that most people interact with Claude) and [Claude Code](https://docs.anthropic.com/en/docs/agents-and-tools/claude-code/overview) (our new specialist coding “agent” that can independently accomplish chains of complex tasks using a variety of digital tools).
+To understand these changes in more detail, we conducted an analysis of 500,000 coding-related interactions across [Claude.ai](http://claude.ai/redirect/website.v1.33a1fedf-5747-4b68-9502-0e04849443fe) (the “default” way that most people interact with Claude) and [Claude Code](https://docs.anthropic.com/en/docs/agents-and-tools/claude-code/overview) (our new specialist coding “agent” that can independently accomplish chains of complex tasks using a variety of digital tools).
 
 We found three key patterns:
 

diff --git a/content/blog/research/in-context-learning-and-induction-heads.md b/content/blog/research/in-context-learning-and-induction-heads.md
@@ -141,7 +141,7 @@ Our views on the AI competition between the US and China.
 
 ### Terms and policies
 
-Privacy choices*   [Privacy policy](https://www.anthropic.com/legal/privacy)
+*   [Privacy policy](https://www.anthropic.com/legal/privacy)
 *   [Consumer health data privacy policy](https://www.anthropic.com/legal/consumer-health-data-privacy-policy)
 *   [Responsible disclosure policy](https://www.anthropic.com/responsible-disclosure-policy)
 *   [Terms of service: Commercial](https://www.anthropic.com/legal/commercial-terms)

diff --git a/content/blog/research/interpretability-dreams.md b/content/blog/research/interpretability-dreams.md
@@ -3,8 +3,6 @@ Title: Interpretability Dreams
 URL Source: https://www.anthropic.com/research/interpretability-dreams
 
 Markdown Content:
-# Interpretability Dreams \ Anthropic
-
 [Skip to main content](https://www.anthropic.com/research/interpretability-dreams#main-content)[Skip to footer](https://www.anthropic.com/research/interpretability-dreams#footer)
 
 [](https://www.anthropic.com/)
@@ -160,3 +158,5 @@ Privacy choices*   [Privacy policy](https://www.anthropic.com/legal/privacy)
 *   [](https://www.linkedin.com/company/anthropicresearch)
 *   [](https://x.com/AnthropicAI)
 *   [](https://www.youtube.com/@anthropic-ai)
+
+# Interpretability Dreams \ Anthropic
diff --git a/content/blog/research/mapping-mind-language-model.md b/content/blog/research/mapping-mind-language-model.md
@@ -21,7 +21,7 @@ There was both an engineering challenge (the raw sizes of the models involved re
 
 As for the scientific risk, the proof is in the pudding.
 
-We successfully extracted millions of features from the middle layer of Claude 3.0 Sonnet, (a member of our current, state-of-the-art model family, currently available on [claude.ai](https://claude.ai/redirect/website.v1.b1ef2abc-07bf-4def-a05d-35e4a35fc8c1)), providing a rough conceptual map of its internal states halfway through its computation. This is the first ever detailed look inside a modern, production-grade large language model.
+We successfully extracted millions of features from the middle layer of Claude 3.0 Sonnet, (a member of our current, state-of-the-art model family, currently available on [claude.ai](https://claude.ai/redirect/website.v1.6b08c2cb-d676-49ea-a3d5-4d257df65e79)), providing a rough conceptual map of its internal states halfway through its computation. This is the first ever detailed look inside a modern, production-grade large language model.
 
 Whereas the features we found in the toy language model were rather superficial, the features we found in Sonnet have a depth, breadth, and abstraction reflecting Sonnet's advanced capabilities.
 

diff --git a/content/blog/research/natural-language-autoencoders.md b/content/blog/research/natural-language-autoencoders.md
@@ -53,7 +53,7 @@ For example, in the blackmail test discussed above, we used NLAs to understand w
 
 Claude does not say it thinks it’s being tested in this transcript from our blackmail evaluation, but the NLA explanation reveals its suspicions.
 
-Extending this analysis to other evaluations, NLAs suggest that Claude suspects it’s being tested more often than it lets on. For instance, in a test of whether Claude takes destructive actions while writing code—for example, deleting important files—NLA explanations show signs of evaluation awareness 16% of the time, even though Claude never explicitly verbalizes this. Similarly, on the popular coding benchmark SWE-bench Verified, which we use to determine how capable Claude is, NLAs indicate that Claude is aware it is being evaluated on 26% of problems. In contrast, NLA explanations indicate evaluation awareness on less than 1% of real [claude.ai](http://claude.ai/redirect/website.v1.6518b370-b0b5-4a69-8954-bfa93edc4441) usage that opted in for training.
+Extending this analysis to other evaluations, NLAs suggest that Claude suspects it’s being tested more often than it lets on. For instance, in a test of whether Claude takes destructive actions while writing code—for example, deleting important files—NLA explanations show signs of evaluation awareness 16% of the time, even though Claude never explicitly verbalizes this. Similarly, on the popular coding benchmark SWE-bench Verified, which we use to determine how capable Claude is, NLAs indicate that Claude is aware it is being evaluated on 26% of problems. In contrast, NLA explanations indicate evaluation awareness on less than 1% of real [claude.ai](http://claude.ai/redirect/website.v1.bc7e7e7d-f778-4a22-9852-5b3feb477007) usage that opted in for training.
 
 ![Image 4](https://www.anthropic.com/_next/image?url=https%3A%2F%2Fwww-cdn.anthropic.com%2Fimages%2F4zrzovbb%2Fwebsite%2Ffc08005f32ff020f21e598cc72f58768af2bcc4f-1590x1170.png&w=3840&q=75)
 

diff --git a/content/claude-code-manifest.json b/content/claude-code-manifest.json
@@ -1,12 +1,12 @@
 {
   "name": "@anthropic-ai/claude-code",
-  "version": "2.1.154",
+  "version": "2.1.156",
   "author": {
     "name": "Anthropic",
     "email": "support@anthropic.com"
   },
   "license": "SEE LICENSE IN README.md",
-  "_id": "@anthropic-ai/claude-code@2.1.154",
+  "_id": "@anthropic-ai/claude-code@2.1.156",
   "maintainers": [
     {
       "name": "zak-anthropic",
@@ -73,20 +73,20 @@
     "claude": "bin/claude.exe"
   },
   "dist": {
-    "shasum": "46815e4dfae0da4cd7bb49f650f7c3f798e96f39",
-    "tarball": "https://registry.npmjs.org/@anthropic-ai/claude-code/-/claude-code-2.1.154.tgz",
+    "shasum": "32f21eb881e84f421195873842c68c367093d43e",
+    "tarball": "https://registry.npmjs.org/@anthropic-ai/claude-code/-/claude-code-2.1.156.tgz",
     "fileCount": 7,
-    "integrity": "sha512-1LSc7Jtm7Jy/GgqzoC4jDtTeV4GphGdB2+r49QDJnDa4fDsJfxVHR4TQNETeyNtGw/uD1+voo7f2Vhjldyqe8w==",
+    "integrity": "sha512-DRIqsawy+n+LtNBaxOW+3JYLaehbCdEdc+mZjYv/zRnZ1bHeTetJBcV41TagNjL00hHjrlALdl76wmA4s/PVQQ==",
     "signatures": [
       {
-        "sig": "MEUCIFkL6YSLE7iKTkKBeNo3L67P4ZetJz58XixxXhK20rjcAiEApdebuVHImUSYJDmRGP7ZKKGZgdrxE+Qfwgr6wVVKFKg=",
+        "sig": "MEUCIAz7iPG3QRZaIOiHKoXT1QSJwRGEkNW2fLQJeEHoeTU8AiEAgtiITFRAZCekOQ17GEVlOht7C4javPiJRal5Ii4AJDI=",
         "keyid": "SHA256:DhQ8wR5APBvFHLF/+Tc+AYvPOdTpcIDqOhxsBHRwC7U"
       }
     ],
     "unpackedSize": 145690
   },
   "type": "module",
-  "_from": "file:staged-npm/anthropic-ai-claude-code-2.1.154.tgz",
+  "_from": "file:staged-npm/anthropic-ai-claude-code-2.1.156.tgz",
   "engines": {
     "node": ">=18.0.0"
   },
@@ -98,8 +98,8 @@
     "name": "wolffiex",
     "email": "wolffiex@anthropic.com"
   },
-  "_resolved": "/home/runner/work/claude-cli-internal/claude-cli-internal/staged-npm/anthropic-ai-claude-code-2.1.154.tgz",
-  "_integrity": "sha512-1LSc7Jtm7Jy/GgqzoC4jDtTeV4GphGdB2+r49QDJnDa4fDsJfxVHR4TQNETeyNtGw/uD1+voo7f2Vhjldyqe8w==",
+  "_resolved": "/home/runner/work/claude-cli-internal/claude-cli-internal/staged-npm/anthropic-ai-claude-code-2.1.156.tgz",
+  "_integrity": "sha512-DRIqsawy+n+LtNBaxOW+3JYLaehbCdEdc+mZjYv/zRnZ1bHeTetJBcV41TagNjL00hHjrlALdl76wmA4s/PVQQ==",
   "_npmVersion": "11.13.0",
   "description": "Use Claude, Anthropic's AI assistant, right from your terminal. Claude can understand your codebase, edit files, run terminal commands, and handle entire workflows for you.",
   "directories": {},
@@ -108,17 +108,17 @@
   "_hasShrinkwrap": false,
   "readmeFilename": "README.md",
   "optionalDependencies": {
-    "@anthropic-ai/claude-code-linux-x64": "2.1.154",
-    "@anthropic-ai/claude-code-win32-x64": "2.1.154",
-    "@anthropic-ai/claude-code-darwin-x64": "2.1.154",
-    "@anthropic-ai/claude-code-linux-arm64": "2.1.154",
-    "@anthropic-ai/claude-code-win32-arm64": "2.1.154",
-    "@anthropic-ai/claude-code-darwin-arm64": "2.1.154",
-    "@anthropic-ai/claude-code-linux-x64-musl": "2.1.154",
-    "@anthropic-ai/claude-code-linux-arm64-musl": "2.1.154"
+    "@anthropic-ai/claude-code-linux-x64": "2.1.156",
+    "@anthropic-ai/claude-code-win32-x64": "2.1.156",
+    "@anthropic-ai/claude-code-darwin-x64": "2.1.156",
+    "@anthropic-ai/claude-code-linux-arm64": "2.1.156",
+    "@anthropic-ai/claude-code-win32-arm64": "2.1.156",
+    "@anthropic-ai/claude-code-darwin-arm64": "2.1.156",
+    "@anthropic-ai/claude-code-linux-x64-musl": "2.1.156",
+    "@anthropic-ai/claude-code-linux-arm64-musl": "2.1.156"
   },
   "_npmOperationalInternal": {
-    "tmp": "tmp/claude-code_2.1.154_1779982850242_0.32321079600865255",
+    "tmp": "tmp/claude-code_2.1.156_1779999172460_0.1439017309709112",
     "host": "s3://npm-registry-packages-npm-production"
   }
 }
diff --git a/content/en/agents-and-tools/tool-use/overview.md b/content/en/agents-and-tools/tool-use/overview.md
@@ -121,7 +121,7 @@ When you use `tools`, the API also automatically includes a special system promp
 
 | Model                    | Tool choice                                          | Tool use system prompt token count          |
 |--------------------------|------------------------------------------------------|---------------------------------------------|
-| Claude Opus 4.8                | `auto`, `none`<hr />`any`, `tool`   | 290 tokens<hr />410 tokens |
+| <NextOpus />                   | `auto`, `none`<hr />`any`, `tool`   | 290 tokens<hr />410 tokens |
 | Claude Opus 4.7                | `auto`, `none`<hr />`any`, `tool`   | 675 tokens<hr />804 tokens |
 | Claude Opus 4.6              | `auto`, `none`<hr />`any`, `tool`   | 497 tokens<hr />589 tokens |
 | Claude Opus 4.5            | `auto`, `none`<hr />`any`, `tool`   | 496 tokens<hr />588 tokens |