”Accenture caps employee’s Claude spending to $29/month.“ […] ”Additional spending requires to be charged through internal work breakdown structures with proper authorization.“ When people ask me why they should use @OpenClaw over e.g. Claude, the answer is simple: technological sovereignty. With OpenClaw I own the harness, I control the costs, data and what models are used. I can combine, mix and switch models, cloud and/or local. I own my workflows, the routines, the integrations, I control what goes in and out. I can choose where to run it, and how to access it. I don't have to submit to some outside corporate policy or agenda. My agents keep working while you have to wait for an outage to end. 404media.co/accenture-tells-…

Does anyone have experience with this Codex plugin and can recommend it for server infrastructure development? chatgpt.com/plugins/tarot

Adobe subscriptions are one of the biggest scams in the history of software. @Adobe if you force me to choose between CHANGING or CANCELLING my subscription on the 4th(!!) screen in the cancellation process, how about you ENABLE that CANCEL BUTTON as well?

People figuring out electronic devices need electricity. More at 10.

I've reached a point where I try to avoid using Claude Code for projects related to “OpenClaw” whenever possible because as soon as I mention OpenClaw, whether it's for research or part of a spec or just a passing reference in a convo, the number of times Claude suddenly makes “mistakes” or “forgets” to do something it should do, increases noticably. My flavor of tism is pattern recognition and I am not imaging this. Case in point, I'm currently working on a robot firmware and as soon as I started mentioning a corresponding plugin for OpenClaw so that my claw can control that robot, Opus 5.5 high throws away one of the three main features (that already existed): vision. "VLM is missing. It was a decision I had to weigh up whilst building, which I didn’t then tell you about." Oh, really.

If you happen to use OpenCodex and have an ChatGPT Pro 500 subscription but don't see #Ultrafast option for Astra and/or gpt-6.1-sol in the model selector even though you've activated it in #OpenCodex, give the following to Codex: Please add “Ultrafast” as a selectable speed in my existing OpenCodex / Codex Desktop installation for these models: - gpt-6-astra - gpt-6.1-sol I explicitly authorize: - read-only inspection of the relevant installation and settings, - creating a timestamped backup of the active model catalog, - adding the Ultrafast option to the existing native entries for these two models, - verifying the saved changes afterward. Do not restart Codex Desktop automatically or send any model request. I will handle the restart and first test myself. Follow these steps: 1. Identify the actual environment. - Determine the installed OpenCodex version and, where possible through read-only inspection, the running version. - Determine which Codex configuration directory the Desktop app uses. Account for a custom CODEX_HOME or another explicitly selected configuration. - Read the model_catalog_json setting from the active Codex configuration. - Use exactly the catalog file referenced by that setting. Do not reuse user-specific paths from another computer. If model_catalog_json is missing, the referenced file does not exist, or the Desktop app uses a different catalog, do not make speculative changes. Explain the finding and the next necessary step. 2. Inspect the existing model entries and settings. - Locate the existing entries with these exact IDs: gpt-6-astra gpt-6.1-sol - Check that service_tiers is a list of objects and additional_speed_tiers is a list of strings. - Check whether ultrafast is already present. - Inspect the relevant OpenCodex fastMode setting. In the OpenCodex version previously inspected here: - fastMode = true forces regular Fast/priority. - fastMode = false removes the requested service tier. - An absent fastMode setting passes through the tier requested by the client. Verify this behavior for the installed version if such an override is present. Do not silently change fastMode or other routing settings. The OpenCodex ultraFastTier switch does not automatically add Ultrafast to the native model picker. An enabled switch therefore does not prove that the catalog is already configured correctly. If these models exist only with provider or account prefixes, investigate how those catalog entries are handled first. Do not create new models or copy complete model entries speculatively. Do not read credentials or print tokens, cookies, or authentication files. 3. Create a backup before making any changes. - Back up the complete active catalog file in the same directory. - Use a new, unique filename, for example: opencodex-catalog.json.backup-YYYY-MM-DD_HH-MM-SS - Do not overwrite an existing backup. - Verify through a byte comparison or SHA-256 that the backup exactly matches the original. - Stop if the backup cannot be created or verified. If all requested entries are already fully present, no additional backup or modification is necessary. 4. Add only the missing Ultrafast entries. In service_tiers for each existing model, add: { "id": "ultrafast", "name": "Ultrafast", "description": "Lower latency, increased usage; requires an eligible plan" } In additional_speed_tiers for each existing model, add: "ultrafast" Observe these constraints: - Preserve all existing speed options. - Avoid duplicate entries. - If one of the two fields already contains the correct entry, modify only the other field. - Preserve slug, display_name, context windows, reasoning settings, and all other fields. - Do not replace the complete model entry with a template. - Do not modify other models. - Do not change the global service_tier default in config.toml. - Do not change credentials or account assignments. Write the file atomically where possible: first write and validate a complete temporary file in the same directory, then replace the catalog file. Before replacing it, verify that the original file has not changed concurrently since it was read. If it has changed, abort the write and inspect the new state. 5. Verify the result. - Read the saved file again. - Confirm that it is valid JSON. - Confirm that the intended models contain ultrafast in both fields. - Compare the saved state with the backup. - Confirm that the only changes are the intended additions to service_tiers and additional_speed_tiers. - Record the change, with a date, in a short local changelog in the working directory. Do not claim that a restart, a request, or actual processing with Ultrafast succeeded unless you verified it. 6. Report the following information: - The full path of the modified catalog file. - The full path of the verified backup. - The specific models and fields that were modified. - The results of the JSON and change verification. - A copyable restore command using the actual, properly quoted absolute paths. - The manual steps for restarting Codex Desktop and selecting Ultrafast. Do not execute the restore command.