OpenAI patched a Codex defect in June that was silently burning through SSD write endurance, but Mac users are again reporting sluggish performance that outlasts the application itself. A fresh wave of complaints points to a macOS Gatekeeper process—syspolicyd—that gets stuck in a loop, demanding a full system restart to restore responsiveness.
What’s new in this wave of reports
The latest round of trouble surfaced on July 18, 2026, in a Reddit thread that quickly gathered over 100 upvotes. Users described their MacBooks becoming broadly unresponsive while Codex was open or even idling in the background. Symptoms included browser stalls, delayed typing, erratic mouse tracking, stuttering graphics, and painfully slow app launches. Several posters said the only reliable fix was to reboot macOS.
Crucially, the slowdown didn’t always end when Codex was quit. In many cases, the machine remained sluggish until a restart, suggesting a system-level component rather than the app itself was the culprit. A second Reddit discussion soon confirmed similar patterns, with some Windows users chiming in about their own desktop-client lag, freezes, and mouse stutter—though the underlying cause on Windows remains less clear.
These anecdotes align with an open GitHub issue (#25719) first filed on June 1. The reporter there shows the Codex desktop app repeatedly triggering a runaway in macOS’s Gatekeeper daemon: syspolicyd spiking to 125–200% CPU (that’s more than one full core) and swelling past 8 GB of memory the moment Codex launches. Because syspolicyd is a system service that validates every app you open, a wedged instance drags the whole machine down—even after Codex and its helper processes exit.
At the same time, the original SSD-wear concern hasn’t been fully laid to rest. GitHub issue #29876, filed just two days after the June 22 logging fix, documented a MacBook Air with an M4 chip still sustaining roughly 207 MB per minute of writes while Codex sat mostly idle. The reporter also found a code_sign_clone cache folder that had ballooned to 12 GB. That’s a far cry from the earlier 37 TB over 21 days, but it’s still abnormal churn on a machine with soldered storage.
Why Macs slow down even after you quit Codex
The persistent lag tracks back to syspolicyd, a core macOS process that enforces Gatekeeper’s code-signing policies. When an application launches a helper, updates a component, or even probes a notarized binary, syspolicyd may validate that code. Under normal circumstances, the check completes quickly and silently. In the case reported on GitHub, however, the daemon gets stuck repeating validation attempts, burning CPU and memory long after the triggering app disappears from the Dock.
This explains the Reddit accounts almost exactly. If Codex’s built-in “Computer Use” helper, its updater, or its background services interact with macOS security policy in a way that trips up syspolicyd, the entire system suffers. Opening unrelated apps becomes slower; graphics and input responsiveness degrade as the OS competes for resources with a runaway policy daemon.
Importantly, the Codex Computer Use helper is signed and notarized, so this isn’t a simple case of an unsigned binary. Instead, the pattern suggests that the way Codex launches, updates, or repeatedly validates its components causes a pathological interaction with Gatekeeper. Some users have reported that the loop occurs even when they disable those features in the config, pointing to a deeper integration snag.
How we got here: from excessive logging to Gatekeeper loops
The original SSD-wear alarm rang in June through GitHub issue #28224. Codex’s local diagnostic logger was ignoring the RUST_LOG environment variable and writing at TRACE-level verbosity—the noisiest possible setting—to a SQLite database at ~/.codex/logs_2.sqlite. Apache Flink PMC member Rui Fan measured the damage: after about 21 days of uptime, the main SSD had written roughly 37 TB, extrapolating to an annualized rate that would consume a typical consumer drive’s entire write-endurance budget in under a year.
OpenAI responded with the 0.142.0 release on June 22, 2026. The fix cut log volume by approximately 85% by the original reporter’s own measurement. Yet as noted, follow-up issue #29876 showed that some systems still experienced anomalous write activity. Around the same time, issue #25719 was gathering detail on the syspolicyd problem, though it hadn’t yet drawn wide notice. The July 18 Reddit surge brought that Gatekeeper loop into the spotlight, demonstrating that the logging fix alone wasn’t enough to fully tame Codex’s impact on Mac hardware.
What you should do right now
First, keep calm—there are no confirmed cases of drives failing outright, but the wear is cumulative and no one wants to burn through their SSD’s lifespan unnecessarily. For the performance lag, the steps below will help you diagnose whether your system is affected and which problem you’re facing.
1. Update Codex.
Whether you’re running the standalone Codex app or the ChatGPT desktop client that now delivers the Codex experience, make sure you’re on the latest build. The 0.142.0 logging fixes are worth having even if they don’t resolve every current complaint.
2. Diagnose with Activity Monitor.
Open Activity Monitor before launching Codex, and note baseline CPU and memory usage. Then launch Codex, use it normally, and after exiting, watch the monitor for a few minutes. Search for these processes:
- syspolicyd
- trustd
- Codex
- Any Codex-related helpers or “Computer Use” processes
If syspolicyd stays pegged at high CPU or memory after Codex is gone, you’ve likely hit the Gatekeeper loop. A reboot is the most reliable way to clear it. Do not force-kill system policy daemons unless you’re an administrator comfortable with potential side effects.
3. Check your SSD write totals.
What matters is cumulative host writes over time, not a single spike. On macOS, you can use smartctl (install via Homebrew) to query SMART data, though Apple’s hardware and controller reporting can be less straightforward than on a typical Windows PC. On Windows, CrystalDiskInfo exposes the “Total Host Writes” value on many NVMe SSDs. If you see the figure climbing aggressively while Codex is idle, you may still be experiencing elevated write activity.
4. Avoid relying on risky workarounds.
A community workaround redirects ~/.codex/logs_2.sqlite to a RAM-backed path via a symlink. That can reduce flash writes, but it also means a crash or reboot discards diagnostic data, and it changes a file Codex expects to be persistent. It is a stopgap for technically confident users, not a substitute for an OpenAI fix. Moreover, symlinking won’t help if your lag is caused by syspolicyd rather than SQLite churn.
5. If your Mac becomes unresponsive, reboot.
Quitting Codex and restarting macOS remains the low-risk recovery path. It’s inconvenient, but it clears the stuck daemon and restores normal performance.
For Windows users: you don’t have Gatekeeper’s syspolicyd, so a matching root cause is unlikely. However, the same principles apply: monitor CPU, memory, and disk activity (via Task Manager or Process Explorer) for Codex and its child processes, and watch for persistent background activity. The recent lag reports on Windows deserve the same scrutiny.
The outlook: what OpenAI needs to address
As of July 19, OpenAI hasn’t publicly commented on the renewed performance complaints. GitHub activity shows ongoing fixes and discussions around Codex state, logging, and desktop behavior, but there’s no official timeline for a syspolicyd-related patch.
The resolution will likely require OpenAI to reproduce and isolate the Gatekeeper interaction. Until then, affected Mac users should treat a reboot as an unfortunate but necessary part of the Codex workflow. For everyone else, staying current with updates and keeping an eye on system monitors is the best defense.
Codex and similar AI coding agents are no longer just cloud services—they’re desktop software with local databases, caches, helper binaries, and long-lived background processes. The operational lessons are the same as for any heavy developer tool: watch process lifetimes, disk-write rates, cache growth, and post-exit cleanup. The next few updates will tell whether OpenAI can turn Codex from a system stressor into a seamless companion.