XIO

xWire

Posts from every feed QUEST has found across the web — blogs, news, podcasts, channels and more.

Cursor - Community Forum - Topics tagged grok

Topics tagged grok

Help Cursor no longer working

Where does the bug appear (feature/product)? Cursor IDE Describe the Bug Cursor repeatedly enters a cycle where it identifies an error, claims to have fixed it, and then either reports the same error again or introduces new errors caused by its attempted fix. It frequently fails to follow explicit instructions, makes changes beyond the requested scope, and modifies code that was previously working. Instead of diagnosing the root cause, it applies successive patches that compound the original…

Expensive Tasks after toppping up

Why does a single task cost over a $1 after going into top up credits situatuation? I spent $5 on bugbot for this same feature. It was a normal feature that I might do once per day in the past. I use grok 4.7 on regular speed. this is my first week after unifying my account. The marketing promised more grok usage which is fine with me because I only use grok 4.7. my normal grok bot routines consumed 77 percent of my usage and a few cursor features ate up the rest of my budget 4 days early. I…

Composer 2.5 Quality Regression — Unrequested Changes, Incomplete Tasks and Increased Costs

Where does the bug appear (feature/product)? Cursor IDE Describe the Bug Versão 1 — Português brasileiro (PT-BR) - Título: Queda significativa na qualidade do Composer 2.5 e problemas recorrentes com Grok 4.6 e 4.7 Gostaria de registrar minha insatisfação com a qualidade dos modelos disponíveis no Cursor nas últimas semanas, principalmente com o Composer 2.5, que anteriormente era meu modelo preferido e apresentava excelentes resultados. Utilizo o Cursor diariamente para desenvolver um sistema…

Cursor Agent request ends with “Cancellation token requested”

Describe the Bug While using Cursor Agent, a request ended with this error: RetriableError: [internal] Cancellation token requested for stream The request did not return a normal response. This happened while [briefly describe what you were asking Cursor to do]. It has happened [once / several times]. [If applicable: I did not manually cancel the request or start another request.] Steps to Reproduce Opened my workspace in Cursor. Started an Agent request to [briefly describe the task]. The…

Failed to generate commit message: [resource_exhausted] Error xxxxx

Where does the bug appear (feature/product)? Cursor IDE Describe the Bug Failed to generate commit message: [resource_exhausted] Error i keep this isse all the time disregards version upgrade with random breaks when its really working. Also when this occasionally works the rules have set that it should follow the semantics of the branch name are not honored Steps to Reproduce generate commit message button Expected Behavior commit generation works ? Operating System MacOS Version Information…

Custom subagent `model:` frontmatter ignored; Task tool only offers "inherit" + parent model (regression since ~Oct 2026)

Where does the bug appear (feature/product)? Cursor IDE Describe the Bug Every custom subagent in .cursor/agents/ runs on the parent chat’s model, whatever its model: frontmatter says. The Task tool’s model parameter only offers inherit and the parent’s own model slug. With an Opus 5.5 parent it offers inherit, claude-opus-5-5-high; with a Sonnet 5.5 parent, inherit, claude-sonnet-5-5. So there’s no way to run a subagent on a different model. Frontmatter values tested, all ignored (subagent ran…

After the acquisition, Cursor is turning into a use-it-and-toss-it accessory

Honestly, it took me a while to write this. I’ve been on Cursor for a long time — Tab completion, Chat, Agent, the whole thing. I renewed my annual plan more than once without thinking twice. On the day of the acquisition I told myself it might actually be good: a bigger parent company, more resources, right? Looking back, it feels more like Cursor got turned into a battery. The charge got drained into Grok, and once it’s empty, you toss it. The signs are pretty clear. Tiers got cut, benefits…

I subscribed to X Ultra last night. I canceled renewal this afternoon

I subscribed to X Ultra to use Cursor and Grok Bot. Getting the Blue Check on my company’s X (Twitter) account was a nice touch (I created an X account for my company because it would benefit from this subscription), and I really would have used Grok more, but instead I will be canceling everything related to Grok. Cursor’s greatest feature was the ability to choose which model I use and avoid Grok models for important, large tasks. Instead, I found out that Cursor does not support external…

Is Cursor Losing Sight of What Made It Great? A Concern From a Long-Time Customer

I’ve been thinking about writing this for a while, and after the recent security incident and subsequent updates, I feel it’s time to share some honest feedback. First, a small disclaimer: This post was written with the help of AI, which did a great job of putting my growing frustration into words. The opinions and experiences behind it are entirely my own. My team and I have been using Cursor since its early days. We genuinely loved the product, believed in its vision, and were excited to see…

" /subagent" invocation never calls Task and repeats an unrelated tool sequence

Where does the bug appear (feature/product)? Cursor CLI Describe the Bug Where: Cursor CLI agent (cursor 2026.11.13-eacade1, Linux 5.4) Model: Grok 4.7 256K High Fast (parent). The custom subagent is .cursor/agents/plan-review.md (model claude-sonnet-5-5). The subagent never starts. Steps to Reproduce In a new chat, run /subagent-plan-review with a new task (read a Jira issue and plan a CLI change). Cursor rewrites the turn as: Use the Task tool with subagent_type=“plan-review” to: Expected:…

Cloud Desktop inaccessible — Claude Sonnet usage limit despite Grok 4.6 selected

Describe the Bug Hello Cursor Support Team, I need assistance with a persistent Cloud Desktop issue that is blocking my website development. I use Cursor Cloud Agents exclusively through Chrome, without the desktop application. Problem: The Desktop panel displays: “The agent’s machine may not be running. Send a message to restart.” The agent reports that its cloud machine is running and the website is available internally, but the interactive Desktop cannot connect. I am using Grok 4.6 High…

iOS app resets the model picker to Grok 4.7 High Fast before every message

Where does the bug appear (feature/product)? Cursor for iOS Describe the Bug On iPhone, every time I compose the next message in an existing agent session, the model picker is Grok 4.7 High with Fast on. I can switch it to Grok 4.7 Medium, or to Grok 4.7 High with Fast off, and send. The following message in that same session is High Fast again. This happens in every session, including sessions created with another model. On the desktop app, the same session keeps the model last used in that…

Ghost of max mode

Where does the bug appear (feature/product)? Editor, Tab & Chat (autocomplete, Composer, in-editor agent) Terminal & commands Models, pricing & API keys (availability, Auto/Max, BYOK/Bedrock) MCP & tools Cloud Agents & Automations (cursor.com/agents, scheduled/event) BugBot & Code Review Cursor CLI Cursor Mobile Remote (SSH / Dev Containers / WSL) Account, billing & login Something else… Describe the Bug A clear and concise description of what the bug is. The remaining ghost of the already dead…

Screen flickering, sessions frequently interrupted

Describe the Bug Lately, while using the Cursor IDE, my sessions have been stopping automatically for no apparent reason—often getting stuck and remaining unresponsive for long periods. Each time this happens, I have to manually prompt the process to resume. Furthermore, I frequently encounter an issue where the entire application screen turns black and all UI components disappear; after about two seconds, the interface returns, but all active sessions have been terminated, forcing me to…

Cursor Automatically Switches from Grok 4.6 to GPT-5.6 sol During Build

Where does the bug appear (feature/product)? Cursor IDE Describe the Bug When I set a plan to use Grok 4.6, Cursor automatically switches to using GPT-5.6 sol instead, even though I explicitly specified Grok 4.6. Steps to Reproduce 1 - I chatted with Claude Opus 5.5 and created a plan. 2 - I opened the plan and saw that Grok 4.6 was already pre-selected to execute it, so I clicked “Build” to begin the process. 3 - Once the build was complete, I checked the plan and usage, and noticed that the…

Subagent model: frontmatter ignores reasoning effort (Grok 4.6 always runs High)

Aqui vai o texto para o fórum, em inglês. Ele não cita o seu nome nem o do projeto. Title: Subagent model: frontmatter ignores reasoning effort (Grok 4.6 always runs High) Hi all, I’m building a local plugin with custom subagents (agents/*.md) and I’m trying to pin each subagent to a model and a reasoning effort. The model part works, but the effort seems to be ignored: the subagent card always shows Grok 4.6 High. What I tried in the agent frontmatter: --- name: qa description: Browser QA…

Repeated Failures Despite Clear Instructions and Visual References

Where does the bug appear (feature/product)? Cursor IDE Describe the Bug I spent about 4% of my Cursor usage quota on a single agent task. There has been no meaningful progress, and the result is still wrong. Task: In an existing Blender campus script, add a 100m straight on the west side of the oval track. I later drew the exact lines directly in the Blender viewport. The new lines must be straight and parallel along the western straight section. They should overlap the existing white lane…

Cursor models getting stupid and almost unusable recently

I am not satisfied with Cursor’s in-house models, whether it’s the Auto mode or other Grok models. Especially in the past few days, they’ve become so shallow and stupid that I can’t count on them to perform complicated and long tasks. And their answering style has become lazy and over simplified, even if I asked them to think thoroughly. There are too many flaws in their plan. If I happen to spot one, they would just admit and apologize. They would only be able to correct that one thing that…

Grok 4.7 - Transcription tips - I want "one" to be transcribed as "1"

Hi everyone, Is there a way to configure Grok Transcriber to output spoken numbers as digits rather than words? Currently, when audio contains spoken numbers (for example, a user saying “one seven seven seven three zero three…”), the transcriber outputs spelled-out words (“one, seven seven seven…”) instead of digits (“1777 303 303”). This makes it difficult to automatically search, format, or extract booking references, Ticket IDs, and phone numbers. We need spoken numeric data to consistently…

Legacy Cursor Ultra: has your Other Models overflow pool become smaller since November?

I’m on Legacy Cursor Ultra, and I am wondering if anyone has noticed a decrease in Other Models quota relative to Cursor Models since November. Since November, this overflow pool seems to have shrunk considerably. My current measurements suggest its full allowance is now below 5% of the Cursor Models allowance, after accounting for the different token costs. Has anyone else noticed this? Below are my measurements and calculations so others can compare or identify anything I’ve missed. 1.…

Something’s changed. Auto mode and Grok. Suddenly they’ve become dumb and lazy

I really enjoyed the work style with Cursor. Usually I use auto and sometimes I would use Grok with extra high effort. Ultra by the way. I loved discussing all the details with Cursor. I used to have the feeling it totally gets me. And it’s so cool so reliable. It could be fast and efficient and it also could be detailed and thoughtful as needed. So auto always my first choice. However yesterday afternoon same conversation continued it just suddenly felt different. The output format became much…

Cloud agents default model setting switches to fast mode

Where does the bug appear (feature/product)? Cursor IDE Describe the Bug i use grok 4.7 for cloud agents, and sometimes i notice it switches to fast mode, why this happens ? Steps to Reproduce i dont know Expected Behavior if i choose a default model, it shouldn’t change and agent should respect to that if i didn’t choose another model Operating System MacOS Version Information Version: 3.28.7 VS Code Extension API: 1.128.0 Commit: 25b52e7d03ebf2d5608914158e119f7992432a30 Date:…

Very slow model

Where does the bug appear (feature/product)? Grok Bot Describe the Bug Ran a very simple task (one prompt) for Grok 4.7 medium fast-mode, to refactor some lines from a React component into a new file (there’s a SKILL file that explains what to do). very simple task. took many long minutes. Request ID: 5d13f75d-ed23-4959-8643-ece9ab876a53 Steps to Reproduce Expected Behavior to be fast not slow. Operating System Windows 10/11 Version Information Version: 3.28.7 (system setup) VS Code Extension…

Reduced performance from Cursor Grok 4.7

I’ve noticed reduced performance from Cursor Grok 4.7. Giving it fairly straightforward tasks, it seems to get stuck thinking for a long time, even though it’s set on “High Fast”. I’ve also seen it produce incomplete results, leaving some methods stubbed, like it is doing the bare minimum. This is in contrast to when it first came out, I was impressed by its speed and thoroughness. Switching to an Anthropic model usually produces a quicker and more thorough solution. But I love the value of the…

Only Grok 4.6 and 4.7 models available

Where does the bug appear (feature/product)? Cursor IDE Describe the Bug Trying to use Composer model but only Grok 4.6 and 4.7 models available. Friday it was available, only change was Cursor restart during the weekend. Steps to Reproduce Models window not showing composer and not able to browse or search Expected Behavior Composer model to be available for selection Screenshots / Screen Recordings Operating System Windows 10/11 Version Information Version: 3.27.22 (user setup) VS Code…

Bug: Surface-owned background shell wakeup drops subagent modelId and falls back to grok-4.7-high-fast (host-owned path preserves lastTurnModel)

Where does the bug appear (feature/product)? Cursor IDE Describe the Bug Summary When a background subagent is spawned via Task with an explicit model override (e.g. model: "grok-4.7-medium", run_in_background: true), its primary turn and any explicit Task(resume) turns correctly run on grok-4.7-medium. However, if a background Shell (block_until_ms: 0) inside that subagent finishes or exits, the resulting BACKGROUND_TASK_COMPLETION follow-up turn ("Briefly inform the user about the task…

Let me set Grok 4.7 High, with fast off, as the default model for cloud agents

Feature request for product/service Cloud Agent (GitHub, Slack, Linear,…) Describe the request When a cloud agent starts without a model specified, it uses the account default, and the only Grok 4.7 option for that default is High Fast. There’s no way to make plain Grok 4.7 High the cloud-agent default, even though I can pick it when I launch agents from the Cursor app. Cloud agents default to High Fast, and I can’t choose High as the default. Fast costs twice as much per token, so my cloud…

Cannot switch models from Grok 4.7 — only Grok 4.7 and Auto are selectable

Where does the bug appear (feature/product)? Cursor IDE Describe the Bug I’m on the Pro+ plan and cannot switch from Grok 4.7 to another model. When I click a different model in the model selector, the selection does not change. The only options I can successfully select are Grok 4.7 and Auto. Steps to Reproduce Open the model selector with Grok 4.7 selected. Click any model other than Grok 4.7 or Auto. The selected model remains Grok 4.7. Expected Behavior The selected model should switch to…