โ† All digests
AI Developer Digest

Wed, Nov 11, 2026

7 signals that cleared the gate20 min read
The Signal โ€” start here
Wednesday, November 11: a quiet day for new lab announcements but a productive one for the agent-framework layer. The main story is PydanticAI v2.67.0 landing Fable 5.2 and Mythos 5.2 support just 36 hours after the model launch โ€” alongside Gemini 4 Carbon, Kimi K3 on Bedrock, and OpenAI's new image models, all in a single release. That breadth is notable: v2.67.0 adds more frontier-model names in a single changelog entry than most releases have in a year. The tradeoff is a dense compatibility-notes section โ€” 20+ items, with at least three that are hard breaks: OpenAI continuation without history_digest is now blocked, CapabilityCreation now refuses below 'operator' authority, and native-tool settings moved to provider_settings. Separately, the Anthropic SDKs (Python v2.8.0, TypeScript sdk-v1.8.0) add Claude Tag analytics to the programmatic reporting surface, suggesting Anthropic's enterprise Slack integration is growing in feature depth. GitHub Trending today shows homeassistant-ai/ha-mcp gaining 640 stars โ€” a signal that MCP is spreading from developer tooling into consumer IoT automation contexts.
Must-reads today
1
PydanticAI v2.67.0 [MEDIUM+BREAKING]: Fable 5.2 / Mythos 5.2 / Gemini 4 Carbon support landed; multiple compat breaks for SubAgents durable execution, provider-specific tool settings, and OpenAI continuation chains
2
Anthropic SDK Python v2.8.0 / TypeScript sdk-v1.8.0 [NOTABLE]: Claude Tag analytics now queryable from the SDK; reply_mode added for Slack channel settings

๐Ÿ† Repo of the Day

1

homeassistant-ai/ha-mcp โ€” 87-tool MCP server that lets AI assistants control Home Assistant natively

On Home Assistant 2026.11+: search "HA-MCP" in Settings โ†’ Devices & Services โ†’ Add Integration

(requires HACS for pre-2026.11 HA installs)

After installation, copy the connect URL from the HA-MCP entry's Configure screen

Then add it to Claude Desktop as an MCP server:

Settings โ†’ Developer โ†’ MCP Servers โ†’ Add Server

Paste the connect URL


Breaking Changes

1
โ—Breaking

PydanticAI v2.67.0 โ€” three compat-note items are hard breaks for affected code

What changed
PydanticAI v2.67.0 ships 20+ compatibility notes. The three that are hard breaks (silent failures or new exceptions thrown):

  1. OpenAI continuation blocked without history_digest: openai_previous_response_id continuation is now refused if the referenced response has no history_digest. Code that continues OpenAI Responses API chains without digest validation will raise an error at runtime.

  1. CapabilityCreation now refuses below 'operator' authority: Calls to create capabilities in a directory below 'operator' authority level now raise. Any code that creates capabilities in user-level or agent-level directories without elevating authority will fail.

  1. Native-tool settings moved to provider_settings: Provider-specific native-tool fields that were passed inline must now go under provider_settings. Existing code using the old shape sends malformed requests silently or raises, depending on the provider.

TL;DR
Three PydanticAI v2.67.0 compat notes that require code changes: OpenAI continuation without history_digest raises, CapabilityCreation authority enforcement is now strict, and native-tool provider settings moved to provider_settings.
Developer signal
Before upgrading to v2.67.0, grep your codebase for openai_previous_response_id โ€” any continuation chain must include a history_digest; add one or catch the new error. Grep for CapabilityCreation and audit the directory argument against your authority model โ€” user-level directory paths need to be elevated to 'operator'. For native-tool settings, move any inline provider-specific fields (e.g., cache_control on Anthropic tools, parallel_tool_calls on OpenAI) under provider_settings. The pydantic-ai changelog uses the phrase "compatibility note" for all 20+ items; not all require action โ€” these three do.


Affects you ifYou continue OpenAI Responses API sessions across turns using openai_previous_response_id; you create capabilities in non-operator directory paths; you pass provider-specific fields to native tools inline.EffortModerate (audit three code patterns, update each)

Model Releases

No new model releases in the 24-hour window. Mistral Large 4 OSS (Oct 29, confirmed) remains unverifiable โ€” mistral.ai DNS-blocked from this session (11th consecutive day).


API & SDK Changes

1
Notable

Anthropic SDK Python v2.8.0 / TypeScript sdk-v1.8.0 โ€” Claude Tag analytics and reply_mode added

What changed
v2.8.0 (Python) and sdk-v1.8.0 (TypeScript), both Nov 10, add:

  • Claude Tag analytics: analytics summaries and users reports now include Claude Tag metrics โ€” usage and activity from Claude in Slack integrations.

  • reply_mode on Claude Tag Slack channel settings: configure how Claude Tag replies in a Slack channel (inline, thread, or ephemeral).

  • Bedrock/Vertex/Google Cloud URL validation: all three platform SDKs now refuse malformed model IDs, project IDs, region strings, or workspace IDs that contain path-traversal sequences (., .., ;) rather than writing them into request paths. Previously, a misconfigured project ID like .. would silently produce an incorrect API path.

TL;DR
Anthropic SDKs (Python v2.8.0 / TypeScript sdk-v1.8.0) add Claude Tag analytics to the programmatic reporting surface and add reply_mode to Claude Tag Slack channel configuration; also harden Bedrock/Vertex/Google Cloud against path-traversal in configuration values.
Developer signal
pip install --upgrade anthropic / npm install @anthropic-ai/sdk@latest to get v2.8.0 / sdk-v1.8.0. If you use the Admin API's analytics endpoints, Claude Tag usage now appears in the summaries and users reports without any change to your query. If you configure Claude Tag Slack channels programmatically, the new reply_mode field lets you set channel-level reply behavior. The URL validation fix is security-relevant: if you're setting Bedrock/Vertex/Google Cloud configuration from user-supplied strings (org ID, region, model), upgrade now โ€” the old code silently wrote malformed values into request paths. The fix raises an AnthropicError immediately instead.


Affects you ifYou query the Admin API analytics endpoints; you configure Claude Tag Slack channels via the API; you set Bedrock/Vertex/Google Cloud configuration from dynamic or user-supplied values.EffortQuick (drop-in version bump; Claude Tag analytics appear automatically)
GitHub (anthropics/anthropic-sdk-python, anthropics/anthropic-sdk-typescript) | Date: 2026-11-10 | Link: https://github.com/anthropics/anthropic-sdk-python/releases/tag/v2.8.0https://github.com/anthropics/anthropic-sdk-python/releases/tag/v2.8.0

Research

arxiv.org DNS-blocked (tenth consecutive day). huggingface.co/papers DNS-blocked (eleventh consecutive day). No research items passed the quality gate from primary sources.


Tooling

2
Medium

PydanticAI v2.67.0 โ€” Fable 5.2, Gemini 4 Carbon, Kimi K3 on Bedrock; KeenableSearch; DBOS child workflows

What changed
v2.66.0 โ†’ v2.67.0: adds Fable 5.2 / Mythos 5.2, Gemini 4 Carbon, Kimi K3 and Kimi K3.1 on Bedrock, OpenAI gpt-image-2.6-flare and gpt-image-2.6-sunburst; ships new KeenableSearch capability (keyless web research), DBOS child workflow tool calls, SubAgents durable execution fix for dynamic_agents, and queue control (dequeue, set_enqueue_priority). Also includes 20+ compatibility notes (three hard breaks in the Breaking Changes section above).
TL;DR
PydanticAI v2.67.0 adds claude-fable-5-2, claude-mythos-5-2, gemini-4-carbon, kimi-k3, and OpenAI image models in a single release, 36 hours after the Fable 5.2 launch; also ships KeenableSearch for keyless web research and DBOS child workflow support.
Developer signal
pip install --upgrade pydantic-ai to get v2.67.0. To use Fable 5.2: model='claude-fable-5-2' or model='claude-mythos-5-2' โ€” both work on direct Anthropic, Bedrock Mantle (gateway/bedrock:global.anthropic.claude-fable-5-2), and OpenRouter. The effort parameter controls thinking depth (replaces budget_tokens โ€” Fable 5.2 follows the Fable 5.2 API conventions covered in the Nov 9 digest). For KeenableSearch: add KeenableSearch() to your agent's capabilities list โ€” it performs web research without requiring an API key, using Claude's built-in search tools. For DBOS child workflows: add child_workflow=True to the DBOSDurability metadata on a tool call; child-workflow tools in the same workspace run in parallel automatically. Read the full compatibility notes before upgrading if you use: SubAgents with dynamic_agents, RepoContext, OpenAI continuation chains, CapabilityCreation, CodeMode, or provider-specific native-tool settings.
Affects you ifYou use PydanticAI as your agent framework and want Fable 5.2 / Mythos 5.2 support; you use DBOS durable execution with SubAgents; you use OpenAI Responses API continuation or provider-specific tool configurations.EffortModerate (review 20+ compat notes before upgrading; three require code changes โ€” see Breaking Changes)
Notable

OpenHands v1.39.0 โ€” automations dashboard owner filter, Agent Server 1.62.0, stop hook feedback isolation

What changed
v1.38.0 (Nov 10) โ†’ v1.39.0 (Nov 11): automations dashboard gains an owner filter with a "Service accounts" option; stop hook feedback is now kept out of the chat (no longer surfaces to end users); chat delivers over HTTP until the session WebSocket is ready (fixes a race condition for slow connections); Agent Server bumped to 1.62.0.
TL;DR
OpenHands v1.39.0 (Nov 11) isolates stop hook feedback from user-visible chat and adds per-owner filtering in the automations dashboard; Agent Server moves to 1.62.0.
Developer signal
The stop hook feedback change is relevant if you use stop hooks for validation or post-processing โ€” previously, any output from a stop hook surfaced in the conversation, which could expose internal prompts or error messages to end users. v1.39.0 routes that feedback to internal logs only. Update with docker pull ghcr.io/all-hands-ai/openhands:1.39.0. If you use the macOS app, clear the quarantine flag after moving to Applications: xattr -dr com.apple.quarantine /Applications/OpenHands.app (the app is not yet notarized โ€” issue #16774).


Affects you ifYou deploy OpenHands with custom stop hooks; you use the automations dashboard with multiple service accounts; you're on macOS using the desktop app.EffortQuick (docker pull or update the desktop app)

Rising Dev Tools

1
Notable

homeassistant-ai/ha-mcp โ€” MCP server that exposes Home Assistant to AI assistants with 87 tools

Developer signal
Works with any MCP client (Claude Desktop, Claude Code, ChatGPT with MCP support, custom agents via the MCP Python SDK). After installation in Home Assistant, copy the connect URL from Settings โ†’ Devices & Services โ†’ HA-MCP โ†’ Configure, and add it as a remote MCP server in your client. Requires a long-lived access token or a dedicated admin account โ€” the owner account is not recommended. Full tool list in the README: ha.list_entities, ha.call_service, ha.create_automation, ha.get_logbook, and 83 more.


Affects you ifYou build agents that interact with smart-home environments; you use Claude Desktop or Claude Code and have a Home Assistant install; you're building home automation test suites that need to verify device states programmatically.
GitHub Trending (all languages, daily, Nov 11) | Stars: ~7.2k (โ†‘ 640 today) | Link: github.com/homeassistant-ai/ha-mcp

Benchmarks & Leaderboards

lmarena.ai DNS-blocked (tenth consecutive day). swebench.com DNS-blocked. No benchmark data confirmed from primary sources this period.


Technical Discussions

simonwillison.net DNS-blocked. hnrss.org inaccessible. Nothing cleared the quality bar this period.


Quick Hits

  • Anthropic SDK TypeScript vertex-sdk-v1.1.3 (Nov 10): sends ; in model IDs as %3B in Vertex request paths; refuses malformed project/region/model strings. Source: github.com/anthropics/anthropic-sdk-typescript/releases/tag/vertex-sdk-v1.1.3
  • Anthropic SDK TypeScript google-cloud-sdk-v1.1.4 / bedrock-sdk-v1.1.2 (Nov 10): same path-traversal validation fix for Google Cloud and Bedrock platform clients. Source: github.com/anthropics/anthropic-sdk-typescript/releases
  • PydanticAI v2.67.0 โ€” Anthropic fast-mode responses now priced at fast rates (#13083): previously, Fable 5.2 fast-mode responses in PydanticAI were billed at standard rates; this is corrected in v2.67.0.
  • Cline / Aider / LangGraph: no releases in 24-hour window (Cline last: Nov 4 nightly; LangGraph last: sdk 0.4.14 Nov 6; Aider last: v0.86.2 Feb 2026).

Worth Watching (Announced, Not Yet Shipped)

  • Gemini API preview model deprecations โ€” November 17 (6 days): Gemini 3.1 Flash TTS preview, Gemini Live API preview models, Gemini 2.5 TTS previews retire. Replacement: gemini-3.8-live and gemini-3.8-flash-tts. Source: ai.google.dev/gemini-api/docs/deprecations (DNS-blocked โ€” verify at official URL).
  • Fast mode removal for Claude Opus 4.8 โ€” November 30 (19 days): speed: "fast" with claude-opus-4-8 will error. Migrate to Claude Opus 5.5. Source: https://platform.claude.com/docs/en/release-notes/overview
  • Claude Sonnet 4.5 retirement โ€” November 30 (19 days): claude-sonnet-4-5-20250929 retires. Migrate to Claude Sonnet 5.5. Source: https://platform.claude.com/docs/en/release-notes/overview


Filtered from 30+ primary sources against a published quality rubric. No press releases, no fluff โ€” only what changes what you build.