{"collection":{"slug":"frontier-ai-labs","name":"Frontier AI Labs"},"digest":{"id":"cwd_1Of49xyQEFk642WV1GvSi","weekStart":"2026-08-10","title":"Speed tiers, browser agents, and the great SDK migration","intro":"The week's defining story is OpenAI's Ultrafast mode for GPT-5.6 Sol, a Cerebras-powered tier that runs up to 14× faster than standard processing. Anthropic countered with Claude Cowork arriving as a Chrome side panel, while OpenAI's Python SDK shipped a breaking HTTPX2 migration.","body":"### The need for speed\n\nOpenAI made speed the headline this week with [Ultrafast mode for GPT-5.6 Sol](/release/rel_y-Z4sa68KDdjXKQApg8L--openai-previews-ultrafast-mode-gpt-5-6-sol-at-up-to-14-speed), previewed first in the API and generating up to 750 output tokens per second — a 14× jump over Standard processing. Powered by Cerebras silicon, the tier targets workflows where latency is the bottleneck: incident response, reliability engineering, anything that can't wait for a typical frontier-model round trip. The [API announcement](/release/rel_FKoDKuzHpmkaYah3iEQIy-openai-api-announces-ultrafast-mode-for-gpt-5-6-sol) confirms it's in limited preview for select customers, with sign-ups open for broader access. It's a notable shift in how frontier labs are competing: not just on benchmark scores but on wall-clock time, treating inference speed as a first-class product feature rather than an infrastructure footnote. The SDKs are already catching up — the [Python SDK v3.1.0](/release/rel_iM7_mtlmqzJNmH79GuIbM-openai-python-sdk-v3-1-0-adds-websocket-stream-ids-and-ultrafast-tier) adds an Ultrafast tier alongside WebSocket stream IDs.\n\n### Agents move into the browser\n\nAnthropic's big play this week was [Claude Cowork as a Chrome side panel](/release/rel_v6LIdhzVwYFdCaILWUd4a-claude-cowork-arrives-as-chrome-side-panel-with-browser-automation), letting Claude read, click, and fill forms on the page you're already signed in to. The notable part is that skills, plugins, and connectors work in the browser for the first time, and sessions sync across desktop and phone — so a task started in a tab can be picked up elsewhere. OpenAI answered on the same front with smaller steps: [Google Drive browsing in Library](/release/rel_BW77HSNQ6nWOfiwRLYfR7-chatgpt-adds-google-drive-browsing-to-library) lets ChatGPT open Docs, Sheets, and Slides beside a conversation and update source files directly, while a new [app directory with connectors as apps](/release/rel_8TwomxzeBz1H3V-2SDnwJ-chatgpt-launches-app-directory-with-connectors-as-apps) consolidates tool management in one place. Neither matches Cowork's reach into the live page, but together they show both labs treating the browser as the next agentic battleground.\n\n### Hardening the CLI and migrating the SDK\n\nClaude Code shipped several releases this week, with [v2.1.232](/release/rel_rnHuDTR8If8c-aU8mYMKe-claude-code-v2-1-232-turns-on-subagent-forking-and-fixes-security-escalations) the most consequential: subagent forking is now on by default, letting `subagent_type: \"fork\"` inherit the full conversation and prompt cache, alongside fixes for PowerShell and Git Bash permission bypasses that could overwrite settings and redirect file access. The follow-up v2.1.233 patches an NTLM credential-leak vector in Windows device-prefix paths and removes todo tools on newer models unless explicitly re-enabled. Meanwhile, the [OpenAI Python SDK v3.0.0](/release/rel_7bbxaDrrfMug3mNgsp6eY-openai-python-sdk-v3-0-0-makes-httpx2-the-default-http-client) is the breaking release of the week: HTTPX2 is now the default HTTP client, httpx is no longer auto-installed, and anyone using custom transports must migrate or drop to a runtime-only legacy escape hatch. If you run a custom HTTPX stack against OpenAI's API, this is the week to read the migration guide.\n\n### Security, enterprise controls, and platform reach\n\nThe rest of the week was steady expansion. [Daybreak cyber models landed on AWS Bedrock](/release/rel_4l41BdiCodKwxODywmRs5-openai-daybreak-cyber-models-now-available-on-aws-bedrock), giving defenders frontier security models inside existing AWS environments. The [ChatGPT desktop app for Linux](/release/rel_gBh-ZMwXaJ04zKbdQCY-l-codex-adds-linux-desktop-preview-and-agent-imports-from-claude-code-and-cursor) hit preview on Ubuntu, Debian, and Fedora, with the notable bonus of importing setup, skills, and recent work from Claude Code and Cursor. Enterprise admins got a wave of controls — audit logs in the Global Admin Console, thread-level Codex cost data, service accounts for automation, and a new [ChatGPT for Intune iOS app](/release/rel_ySf84iAN6cubWJ2bkBnkF-chatgpt-for-intune-ios-app) — while an optional [Computer History feature for macOS](/release/rel_CEDNVrKlP2dNyhZ1cBWYv-chatgpt-adds-optional-computer-history-for-macos-to-resume-context-across-apps) brings cross-app context to Pro, Business, and Enterprise users. The [Go SDK](/release/rel_sZP9KJRSB74LQzBbkfBPh-openai-go-sdk-adds-model-picker-discovery-plugin-url-flags-and-event-callback) also added model-picker discovery and plugin URL flags, and a new [Codex app for macOS](/release/rel_Q2hVb3tKVfvInBv2z_y5V-codex-app-for-macos) targets teams juggling multiple parallel coding agents. A quieter but meaningful change: [project memory settings in ChatGPT](/release/rel_aokxJCD9HvrMdfq6JKj4U-chatgpt-lets-you-edit-memory-settings-for-existing-projects) are now editable after creation, letting teams scope context per project.","releaseIds":["rel_y-Z4sa68KDdjXKQApg8L-","rel_FKoDKuzHpmkaYah3iEQIy","rel_iM7_mtlmqzJNmH79GuIbM","rel_v6LIdhzVwYFdCaILWUd4a","rel_BW77HSNQ6nWOfiwRLYfR7","rel_8TwomxzeBz1H3V-2SDnwJ","rel_rnHuDTR8If8c-aU8mYMKe","rel_7bbxaDrrfMug3mNgsp6eY","rel_4l41BdiCodKwxODywmRs5","rel_gBh-ZMwXaJ04zKbdQCY-l","rel_ySf84iAN6cubWJ2bkBnkF","rel_CEDNVrKlP2dNyhZ1cBWYv","rel_sZP9KJRSB74LQzBbkfBPh","rel_Q2hVb3tKVfvInBv2z_y5V","rel_aokxJCD9HvrMdfq6JKj4U"],"releaseCount":50,"generatedAt":"2026-08-17T06:17:11.656Z","releases":[{"id":"rel_y-Z4sa68KDdjXKQApg8L-","title":"OpenAI previews Ultrafast mode: GPT-5.6 Sol at up to 14× speed","path":"/release/rel_y-Z4sa68KDdjXKQApg8L--ultrafast-tier-runs-gpt-5-6-sol-14-faster","org":{"slug":"openai","name":"OpenAI"},"importance":5},{"id":"rel_FKoDKuzHpmkaYah3iEQIy","title":"OpenAI API announces Ultrafast mode for GPT-5.6 Sol","path":"/release/rel_FKoDKuzHpmkaYah3iEQIy-ultrafast-mode-14x-faster-gpt-5-6-sol-tier","org":{"slug":"openai","name":"OpenAI"},"importance":4},{"id":"rel_iM7_mtlmqzJNmH79GuIbM","title":"OpenAI Python SDK v3.1.0 adds WebSocket stream IDs and Ultrafast tier","path":"/release/rel_iM7_mtlmqzJNmH79GuIbM-websocket-stream-ids-sora-video-apis-deprecated","org":{"slug":"openai","name":"OpenAI"},"importance":3},{"id":"rel_v6LIdhzVwYFdCaILWUd4a","title":"Claude Cowork arrives as Chrome side panel with browser automation","path":"/release/rel_v6LIdhzVwYFdCaILWUd4a-claude-cowork-now-works-in-chrome-side-panel","org":{"slug":"anthropic","name":"Anthropic"},"importance":4},{"id":"rel_BW77HSNQ6nWOfiwRLYfR7","title":"ChatGPT adds Google Drive browsing to Library","path":"/release/rel_BW77HSNQ6nWOfiwRLYfR7-google-drive-files-now-browsable-in-library","org":{"slug":"openai","name":"OpenAI"},"importance":3},{"id":"rel_8TwomxzeBz1H3V-2SDnwJ","title":"ChatGPT launches app directory with connectors as apps","path":"/release/rel_8TwomxzeBz1H3V-2SDnwJ-app-directory-ships-connectors-now-listed-as-apps","org":{"slug":"openai","name":"OpenAI"},"importance":4},{"id":"rel_rnHuDTR8If8c-aU8mYMKe","title":"Claude Code v2.1.232 turns on subagent forking and fixes security escalations","path":"/release/rel_rnHuDTR8If8c-aU8mYMKe-subagent-forking-on-by-default-privilege-escalation-fixes","org":{"slug":"anthropic","name":"Anthropic"},"importance":4},{"id":"rel_7bbxaDrrfMug3mNgsp6eY","title":"OpenAI Python SDK v3.0.0 makes HTTPX2 the default HTTP client","path":"/release/rel_7bbxaDrrfMug3mNgsp6eY-httpx2-becomes-default-custom-clients-must-migrate","org":{"slug":"openai","name":"OpenAI"},"importance":4},{"id":"rel_4l41BdiCodKwxODywmRs5","title":"OpenAI Daybreak cyber models now available on AWS Bedrock","path":"/release/rel_4l41BdiCodKwxODywmRs5-daybreak-blue-and-red-launch-on-aws","org":{"slug":"openai","name":"OpenAI"},"importance":4},{"id":"rel_gBh-ZMwXaJ04zKbdQCY-l","title":"Codex adds Linux desktop preview and agent imports from Claude Code and Cursor","path":"/release/rel_gBh-ZMwXaJ04zKbdQCY-l-linux-desktop-preview-imports-from-claude-code-and-cursor","org":{"slug":"openai","name":"OpenAI"},"importance":3},{"id":"rel_ySf84iAN6cubWJ2bkBnkF","title":"ChatGPT for Intune iOS app","path":"/release/rel_ySf84iAN6cubWJ2bkBnkF-chatgpt-for-intune-ios-app","org":{"slug":"openai","name":"OpenAI"},"importance":null},{"id":"rel_CEDNVrKlP2dNyhZ1cBWYv","title":"ChatGPT adds optional Computer History for macOS to resume context across apps","path":"/release/rel_CEDNVrKlP2dNyhZ1cBWYv-computer-history-for-macos-tracks-app-activity-to-resume-context","org":{"slug":"openai","name":"OpenAI"},"importance":3},{"id":"rel_sZP9KJRSB74LQzBbkfBPh","title":"OpenAI Go SDK adds model picker discovery, plugin URL flags, and event callback improvements","path":"/release/rel_sZP9KJRSB74LQzBbkfBPh-model-picker-now-supports-gateway-discovery","org":{"slug":"openai","name":"OpenAI"},"importance":3},{"id":"rel_Q2hVb3tKVfvInBv2z_y5V","title":"Codex app for macOS","path":"/release/rel_Q2hVb3tKVfvInBv2z_y5V-codex-app-for-macos","org":{"slug":"openai","name":"OpenAI"},"importance":null},{"id":"rel_aokxJCD9HvrMdfq6JKj4U","title":"ChatGPT lets you edit memory settings for existing projects","path":"/release/rel_aokxJCD9HvrMdfq6JKj4U-project-memory-settings-editable-after-creation","org":{"slug":"openai","name":"OpenAI"},"importance":3}]}}