Repository navigation
gpt-6-astra is unusable: cannot finish a simple plan, days-long self-contradicting loops, unrequested changes, ~1 TB written to /tmp #46700
Description
What version of Codex CLI is running?
codex-cli 0.155.1
What subscription do you have?
ChatGPT Pro 20x
Which model were you using?
gpt-6-astra max
What platform is your computer?
Linux 6.6.87.2-microsoft-standard-WSL2 x86_64
What terminal emulator and version are you using (if applicable)?
Windows Terminal (WSL / PowerShell)
Codex doctor report
{
"schemaVersion": 1,
"generatedAt": "1789844985s since unix epoch",
"overallStatus": "ok",
"codexVersion": "0.155.1",
"checks": {
"app_server.status": {
"id": "app_server.status",
"category": "app-server",
"status": "ok",
"summary": "background server is not running",
"details": {
"control socket": "/home/andyalvarez/.codex/app-server-control/app-server-control.sock",
"daemon state dir": "/home/andyalvarez/.codex/app-server-daemon",
"mode": "ephemeral",
"pid file": "/home/andyalvarez/.codex/app-server-daemon/app-server.pid (missing)",
"settings": "/home/andyalvarez/.codex/app-server-daemon/settings.json (missing)",
"status": "not running",
"update-loop pid file": "/home/andyalvarez/.codex/app-server-daemon/app-server-updater.pid (missing)"
},
"remediation": null,
"durationMs": 0
},
"auth.credentials": {
"id": "auth.credentials",
"category": "auth",
"status": "ok",
"summary": "auth is configured",
"details": {
"auth file": "/home/andyalvarez/.codex/auth.json",
"auth storage mode": "File",
"stored API key": "false",
"stored ChatGPT tokens": "true",
"stored agent identity": "false",
"stored auth mode": "chatgpt"
},
"remediation": null,
"durationMs": 0
},
"config.load": {
"id": "config.load",
"category": "config",
"status": "ok",
"summary": "config loaded",
"details": {
"CODEX_HOME": "/home/andyalvarez/.codex",
"active thread overrides": "not inspected",
"config.toml": "/home/andyalvarez/.codex/config.toml",
"config.toml parse": "ok",
"configuration load ms": "14",
"configuration scope": "invocation config, including cloud-managed policy",
"cwd": "/srv/workspace",
"enabled feature flags": "shell_tool, view_image, sleep_tool, unified_exec, unified_exec_tty, unified_exec_zsh_fork, shell_snapshot, content_item_kinds, code_mode_host, terminal_resize_reflow, sqlite, memories, hooks, enable_request_compression, unbounded_connection_retries, multi_agent, apps, tool_search_always_defer_mcp_tools, tool_suggest, plugins, in_app_browser, in_app_chat, in_app_dictation, in_app_local_automation, in_app_updates, browser_use, browser_use_full_cdp_access, browser_use_external, computer_use, remote_plugin, plugin_sharing, image_generation, resize_all_images, item_ids, skill_mcp_dependency_install, skill_search, mentions_v2, steer, guardian_approval, goals, collaboration_modes, tool_call_mcp_elicitation, auth_elicitation, personality, fast_mode, tui_app_server, compaction_image_budget, workspace_dependencies",
"feature flag overrides": "memories=true",
"feature flags enabled": "48",
"log dir": "/home/andyalvarez/.codex/log",
"mcp servers": "1",
"model": "gpt-6-astra",
"model provider": "openai",
"sqlite home": "/home/andyalvarez/.codex"
},
"remediation": null,
"durationMs": 0
},
"git.environment": {
"id": "git.environment",
"category": "git",
"status": "ok",
"summary": "git executable found; execution not verified",
"details": {
"PATH git #1": "/usr/bin/git",
"PATH git #2": "/bin/git",
"PATH git entries": "2",
"git execution": "not inspected (PATH helpers are not executed)",
"repo detected": "false",
"selected git": "/usr/bin/git"
},
"remediation": null,
"durationMs": 52
},
"installation": {
"id": "installation",
"category": "install",
"status": "ok",
"summary": "installation looks consistent",
"details": {
"PATH codex #1": "/home/andyalvarez/.local/bin/codex",
"PATH codex #2": "/usr/bin/codex",
"PATH codex #3": "/bin/codex",
"PATH codex entries": "3",
"current executable": "/home/andyalvarez/.local/lib/node_modules/@openai/codex/node_modules/@openai/codex-linux-x64/vendor/x86_64-unknown-linux-musl/bin/codex",
"install context": "npm (package /home/andyalvarez/.local/lib/node_modules/@openai/codex/node_modules/@openai/codex-linux-x64/vendor/x86_64-unknown-linux-musl, bin /home/andyalvarez/.local/lib/node_modules/@openai/codex/node_modules/@openai/codex-linux-x64/vendor/x86_64-unknown-linux-musl/bin, resources /home/andyalvarez/.local/lib/node_modules/@openai/codex/node_modules/@openai/codex-linux-x64/vendor/x86_64-unknown-linux-musl/codex-resources, path /home/andyalvarez/.local/lib/node_modules/@openai/codex/node_modules/@openai/codex-linux-x64/vendor/x86_64-unknown-linux-musl/codex-path)",
"managed by Vite+": "false",
"managed by bun": "false",
"managed by npm": "true",
"managed by pnpm": "false",
"managed package root": "/home/andyalvarez/.local/lib/node_modules/@openai/codex",
"npm update target": "not inspected (PATH helpers are not executed)"
},
"remediation": null,
"durationMs": 71
},
"mcp.config": {
"id": "mcp.config",
"category": "mcp",
"status": "ok",
"summary": "MCP configuration is locally consistent",
"details": {
"configured servers": "1",
"disabled servers": "0",
"streamable_http servers": "1"
},
"remediation": null,
"durationMs": 239
},
"network.env": {
"id": "network.env",
"category": "network",
"status": "ok",
"summary": "network-related environment looks readable",
"details": {
"managed proxy": "not configured",
"proxy env vars": "none",
"respect system proxy": "disabled"
},
"remediation": null,
"durationMs": 0
},
"network.provider_reachability": {
"id": "network.provider_reachability",
"category": "reachability",
"status": "ok",
"summary": "active provider endpoints are reachable over HTTP",
"details": {
"ChatGPT inference URL": "https://chatgpt.com/backend-api/<redacted> reachable (HTTP 405)",
"reachability mode": "ChatGPT auth"
},
"remediation": null,
"durationMs": 106
},
"network.websocket_reachability": {
"id": "network.websocket_reachability",
"category": "websocket",
"status": "ok",
"summary": "Responses WebSocket handshake succeeded",
"details": {
"DNS": "2 IPv4, 2 IPv6, first IPv4",
"auth mode": "chatgpt",
"connect timeout": "15000 ms",
"endpoint": "wss://chatgpt.com/backend-api/<redacted>",
"handshake result": "HTTP 101 Switching Protocols",
"model provider": "openai",
"provider name": "OpenAI",
"proxy env vars": "none",
"reasoning header": "false",
"server model present": "false",
"supports websockets": "true",
"wire API": "responses"
},
"remediation": null,
"durationMs": 769
},
"runtime.provenance": {
"id": "runtime.provenance",
"category": "runtime",
"status": "ok",
"summary": "running npm on linux-x86_64",
"details": {
"commit": "unknown",
"current executable": "/home/andyalvarez/.local/lib/node_modules/@openai/codex/node_modules/@openai/codex-linux-x64/vendor/x86_64-unknown-linux-musl/bin/codex",
"install method": "npm (package /home/andyalvarez/.local/lib/node_modules/@openai/codex/node_modules/@openai/codex-linux-x64/vendor/x86_64-unknown-linux-musl, bin /home/andyalvarez/.local/lib/node_modules/@openai/codex/node_modules/@openai/codex-linux-x64/vendor/x86_64-unknown-linux-musl/bin, resources /home/andyalvarez/.local/lib/node_modules/@openai/codex/node_modules/@openai/codex-linux-x64/vendor/x86_64-unknown-linux-musl/codex-resources, path /home/andyalvarez/.local/lib/node_modules/@openai/codex/node_modules/@openai/codex-linux-x64/vendor/x86_64-unknown-linux-musl/codex-path)",
"platform": "linux-x86_64",
"version": "0.155.1"
},
"remediation": null,
"durationMs": 0
},
"runtime.search": {
"id": "runtime.search",
"category": "search",
"status": "ok",
"summary": "search command found (bundled); execution not verified",
"details": {
"search command": "/home/andyalvarez/.local/lib/node_modules/@openai/codex/node_modules/@openai/codex-linux-x64/vendor/x86_64-unknown-linux-musl/codex-path/rg",
"search command readiness": "file exists",
"search provider": "bundled"
},
"remediation": null,
"durationMs": 0
},
"sandbox.helpers": {
"id": "sandbox.helpers",
"category": "sandbox",
"status": "ok",
"summary": "sandbox configuration is readable",
"details": {
"approval policy": "OnRequest",
"codex-linux-sandbox helper": "/home/andyalvarez/.codex/tmp/arg0/codex-arg0OMRplK/codex-linux-sandbox",
"denied-read glob rules": "0",
"denied-read rules": "0",
"execve wrapper helper": "/home/andyalvarez/.codex/tmp/arg0/codex-arg0OMRplK/codex-execve-wrapper",
"filesystem sandbox": "restricted",
"glob scan max depth": "unbounded",
"managed filesystem source": "none",
"network sandbox": "restricted"
},
"remediation": null,
"durationMs": 0
},
"security.endpoint": {
"id": "security.endpoint",
"category": "security",
"status": "ok",
"summary": "endpoint protection is not inspected on this platform",
"details": {
"endpoint products": "not inspected on this platform"
},
"remediation": null,
"durationMs": 0
},
"state.paths": {
"id": "state.paths",
"category": "state",
"status": "ok",
"summary": "state paths and databases are inspectable",
"details": {
"CODEX_HOME": "/home/andyalvarez/.codex (dir)",
"active rollout files": "683 files, 30257336929 total bytes, 44300639 average bytes",
"archived rollout files": "0 files, 0 total bytes, 0 average bytes",
"goals DB": "/home/andyalvarez/.codex/goals_1.sqlite (file)",
"goals DB integrity": "ok",
"log DB": "/home/andyalvarez/.codex/logs_2.sqlite (file)",
"log DB integrity": "ok",
"log dir": "/home/andyalvarez/.codex/log (dir)",
"memories DB": "/home/andyalvarez/.codex/memories_1.sqlite (file)",
"memories DB integrity": "ok",
"memories v2 DB": "/home/andyalvarez/.codex/memories_v2_1.sqlite (missing)",
"memories v2 DB integrity": "skipped (missing)",
"queue DB": "/home/andyalvarez/.codex/queue_1.sqlite (file)",
"queue DB integrity": "ok",
"sqlite home": "/home/andyalvarez/.codex (dir)",
"state DB": "/home/andyalvarez/.codex/state_5.sqlite (file)",
"state DB integrity": "ok",
"thread history DB": "/home/andyalvarez/.codex/thread_history_1.sqlite (file)",
"thread history DB integrity": "ok"
},
"remediation": null,
"durationMs": 2151
},
"state.rollout_db_parity": {
"id": "state.rollout_db_parity",
"category": "threads",
"status": "ok",
"summary": "rollout files and state DB thread inventory agree",
"details": {
"default model provider": "openai",
"rollout DB active files": "683",
"rollout DB active rows": "683",
"rollout DB archive mismatches": "0",
"rollout DB archived files": "0",
"rollout DB archived rows": "0",
"rollout DB duplicate DB paths": "0",
"rollout DB duplicate rollout thread ids": "0",
"rollout DB malformed file names": "0",
"rollout DB missing active rows": "0",
"rollout DB missing archived rows": "0",
"rollout DB model providers": "openai=652, mock=30, custom-gateway=1",
"rollout DB rows": "683",
"rollout DB scan cap reached": "false",
"rollout DB scan errors": "0",
"rollout DB sources": "subagent:thread_spawn=406, cli=235, exec=37, subagent:other=2, vscode=2, subagent:review=1",
"rollout DB stale rows": "0"
},
"remediation": null,
"durationMs": 159
},
"system.disk": {
"id": "system.disk",
"category": "disk",
"status": "ok",
"summary": "sufficient free disk space (48.0 GiB)",
"details": {
"CODEX_HOME available": "48.0 GiB",
"failure threshold": "1.0 GiB",
"warning threshold": "5.0 GiB",
"worktree available": "48.0 GiB"
},
"remediation": null,
"durationMs": 0
},
"system.environment": {
"id": "system.environment",
"category": "system",
"status": "ok",
"summary": "OS language en-US",
"details": {
"EDITOR": "not set",
"LANG": "en_US.UTF-8",
"VISUAL": "not set",
"os": "Debian 13.0.0 (trixie) [64-bit]",
"os language": "en-US",
"os type": "Debian",
"os version": "13.0.0"
},
"remediation": null,
"durationMs": 4
},
"terminal.env": {
"id": "terminal.env",
"category": "terminal",
"status": "ok",
"summary": "terminal metadata was detected",
"details": {
"DISPLAY": "present",
"WAYLAND_DISPLAY": "present",
"WSL_DISTRO_NAME": "present",
"WSL_INTEROP": "present",
"WT_SESSION": "present",
"color output": "enabled",
"effective locale": "en_US.UTF-8",
"stderr is terminal": "true",
"stdin is terminal": "true",
"stdout is terminal": "true",
"terminal": "Windows Terminal",
"terminal size": "280x70"
},
"remediation": null,
"durationMs": 0
},
"terminal.title": {
"id": "terminal.title",
"category": "title",
"status": "ok",
"summary": "terminal title default",
"details": {
"terminal title activity": "true",
"terminal title items": "activity, project-name",
"terminal title project source": "cwd",
"terminal title project value": "workspace",
"terminal title source": "default"
},
"remediation": null,
"durationMs": 0
},
"updates.status": {
"id": "updates.status",
"category": "updates",
"status": "ok",
"summary": "update configuration is locally consistent",
"details": {
"cached latest version": "0.155.1",
"check for update on startup": "true",
"last checked at": "2026-09-19T16:46:30.960255820Z",
"latest version": "0.155.1",
"latest version status": "current version is not older",
"npm update target": "not inspected (PATH helpers are not executed)",
"update action": "npm install -g @openai/codex",
"version cache": "/home/andyalvarez/.codex/version.json"
},
"remediation": null,
"durationMs": 272
}
}
}What issue are you seeing?
gpt-6-astra cannot complete a simple, well-specified task at either xhigh or max reasoning effort, and consumes an unreasonable amount of time, quota and disk trying.
The task was a modest change in a Next.js / React / Prisma project. The plan was a markdown file split into verifiable phases, and the repository has an AGENTS.md and skills configured. At xhigh the model worked for more than a full day straight and never got past roughly 30% of the plan.
Observed behavior:
- Non-convergence: it repeatedly reverted and rewrote its own earlier changes, contradicting decisions it had already made. Completed phases were never treated as done.
- Instruction-following failure: it ignored explicit instructions from the plan and from AGENTS.md.
- Scope creep: it never asked a single clarifying question, and created files and changes nobody requested.
- Resource abuse: it wrote close to 1 TB of data to /tmp.
- Quota burn: it consumed 2 of my 3 accumulated usage resets for that 30%.
At max it was worse: the model deleted most of what xhigh had produced, redid it, and fell into the same loop without finishing.
The environment is healthy. codex doctor --json reports overallStatus "ok", the latest CLI version, valid auth, reachable endpoints and a working sandbox, so this points to model behavior and not local configuration.
On a $200/month plan, this is unacceptable. It is also hard to take seriously that this is being marketed as a step toward AGI when the model cannot finish a simple, verifiable plan without rewriting its own work in circles for over a day. Calling this AGI is an embarrassment.
What steps can reproduce the bug?
- Open a Next.js / React / Prisma project with an AGENTS.md and skills configured.
- Provide a markdown plan divided into verifiable phases for a relatively simple task.
- Ask gpt-6-astra to execute it at reasoning effort xhigh, without subagents.
- Let it run and monitor /tmp. The model does not converge, rewrites its own previous work, asks no questions and makes unrequested changes, while /tmp grows to close to 1 TB.
- Repeat the same task at reasoning effort max.
Reproducibility: consistent across multiple attempts at both xhigh and max.
What is the expected behavior?
The model should execute the plan phase by phase, verify each phase, and preserve completed work unless a later step requires changing it. It should follow AGENTS.md and the plan, ask for clarification instead of inventing work, and avoid unrequested changes.
A simple task should finish in hours, not days, with bounded disk usage. A run that is not converging should stop or ask for input instead of looping indefinitely.
Additional information
- Environment: Codex CLI 0.155.1 (npm), WSL2 on Debian 13, Windows Terminal, model gpt-6-astra. The full
codex doctor --jsonoutput is included above, with the username redacted in the paths. - I am a senior full-stack developer with over 15 years of professional experience and a background in electronic engineering. I know how to write a plan, structure a repository and instruct an agent. The task was simple and the plan was verifiable, so this is not a prompting problem.
- For comparison, DeepSeek handles this kind of work considerably better. Its code may be less clean in terms of SRP and DRY, but it delivers something functional. This model delivered 30% of the work, full of contradictions, after more than a day of runtime.
- This looks like a serious regression in reliability, instruction-following and resource control, and it needs an investigation. Other users seeing the same loops, runaway runtimes or /tmp growth are welcome to add their experience below.
- I can share the plan file, the AGENTS.md and the rollout logs on request.
Activity
github-actions commentedon Sep 19, 2026
Potential duplicates detected. Please review them and close your issue if it is a duplicate.
Powered by Codex Action
84dnnvbdvp-debug commentedon Sep 19, 2026
There is one release-specific discriminator for the ~1 TB
/tmpsymptom that may separate Codex-owned temp retention from model-issued file writes.On the exact 0.155.1 tag,
HookOutputSpillerwrites every oversized hook output (default model-visible budget: 2,500 tokens) in full tostd::env::temp_dir()/hook_outputs/<thread_id>/<uuid>.txt, then gives the model a truncated preview plus the saved path. A bounded source search did not find a matchinghook_outputscleanup path. By contrast, ordinary shell-tool stdout/stderr retention is hard-capped (EXEC_OUTPUT_MAX_BYTES = DEFAULT_OUTPUT_BYTES_CAP), and the doctor snapshot in this issue reports rollout files underCODEX_HOME(~30.25 GB), which is a different path from the reported ~1 TB in/tmp.Since the posted config has
hooksenabled, if the existing failed-run temp tree is still present, a read-only attribution check would be useful before deleting anything:If
/tmp/hook_outputsaccounts for most of the growth, that isolates a Codex-owned retention path worth investigating independently of the non-convergence. If it is small or absent, that falsifies this path and the top-levelduoutput should identify which temp subtree actually grew.This does not establish that hooks caused the ~1 TB reported here; it is only a release-specific measurement discriminator.
AndyAlvarezDev commentedon Sep 19, 2026
Thanks for the attribution check. I ran it, with these results:
Those material-integration-* directories are exactly where the growth happened: during those runs /tmp reached more than 700 GB, which I was able to verify at the time. I watched in the CLI as the agent's own test runs created them, repeatedly and without cleaning them up. I no longer have a record of that measurement because /tmp had grown so much that I ran out of disk space, to the point that I couldn't even start a new test run, so I deleted the vast majority of those directories on the spot.
So this is not hook output spilling, but it is still Codex-driven: the agent kept launching test runs that left hundreds of MB of temp data each and never cleaned up after itself.
The main issue is unchanged: the model does not converge (~30% after more than a full day, repeatedly rewriting its own changes).
ruihuabunny commentedon Sep 20, 2026
same problem encountered when using xhigh/max reasoning effort.
AndyAlvarezDev commentedon Sep 21, 2026
This is speculation, and I have no evidence for it, but the behavior (rewriting its own earlier decisions, losing track of completed phases, failing to converge on simple tasks) looks to me like what I would expect from a model served with heavier optimization, such as aggressive quantization, to reduce serving cost. It would also fit with a compute shortage on the provider side, given that the $200 plan has apparently been restricted recently.
ruihuabunny commentedon Sep 21, 2026
I have sent the evidence to OpenAI via the codex cli feedback session. Probably you may try to find the trajectory of codex if you are developing an agentic system. Here is the command:
codex exec --json "your task"
Add corroborated user reports as cautions (not vendor facts)