iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
Jula Markova’s account of an AI content pipeline describes at least 117 intermittent “File does not exist” errors caused, she says, by an ambiguous path convention in YAML briefs. The number is a lower bound from transcript review, not a failure rate or an independently verified count. The incident shows how a successful final run can conceal repeated errors—and why reliable agents need explicit input contracts enforced by both the software that creates their inputs and the agents that consume them.
What the “117 ghost errors” were
In a first-person post published on DEV Community on August 18, 2026, IT analyst Jula Markova recounts auditing session transcripts from an AI content pipeline. She says the audit found at least 117 instances of the same error between May 15 and July 2, across 1,811 transcripts. Markova calls 117 a floor: long transcript messages could be truncated, hiding further occurrences. The count is not an exact total, and it cannot be turned into an error rate because the denominator was transcripts—not path-resolution attempts. Markova’s account is a practitioner incident report, not an independently audited study.
The error line reproduced in the post reads: File does not exist. Note: your current working directory is /Users/userxy/code/your-project. The pipeline still delivered articles, according to Markova, because a failed read was often retried; the agent tried another path and found the file. That made the issue easy to miss if monitoring focused on whether a run ultimately succeeded.
The Tool Desk
Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →Why the same brief could produce different outcomes
Markova attributes the intermittent failures to a YAML brief that mixed absolute paths with a repository-relative article path but did not say what directory relative paths were relative to. A consumer could plausibly resolve that path against its current working directory, the repository root, or the directory containing the brief. If different runs effectively used different interpretations, a file that existed could still be sought in the wrong place.
#1 Best Overall
- EVOLUTION AMD RYZEN AI MAX+ 395 MINI PC - GMKtec EVO-X2 is the next evolution in AI mini PC Ryzen Strix Halo series. Thanks to AMD Simultaneous Multithreading (SMT) the core-count is effectively doubled, to 32 threads. Ryzen AI Max+ 395 has 64 MB of L3 cache and can boost up to 5.1 GHz, depending on the workload. The Ryzen AI Max+ 395 is currently rated as the "most powerful x86 APU" on the market for AI computing.
- AI NPU with XDNA 2 ARCHITECTURE - Powered by 16 “Zen 5” CPU cores, 50+ peak AI TOPS XDNA 2 NPU and a truly massive integrated GPU driven by 40 AMD RDNA 3.5 CUs, the Ryzen AI MAX+ 395 is a transformative upgrade and delivers a significant performance boost over the competition. The Ryzen AI Max+ 395 excels in consumer AI workloads like the llama.cpp-powered application: LM Studio. Shaping up to be the must-have app for client LLM workloads, LM Studio allows users to locally run the latest language model without any technical knowledge required and unleash their creativity and productivity.
- AMD RADEON 8090S iGPU GAMING PC - The AMD Radeon RX 8060S offers all 40 CUs with up to 2.9 GHz graphics clock and uses the new RDNA 3.5 architecture. The powerful iGPU is positioned between an RTX 4060 and 4070 laptop GPU and therefore enables gaming in FHD at maximum details in most demanding games. The 8060S can also utilize the full 128GB pool, which is perfect for running LLMs such as Deepseek 70B Q8, which runs comfortably on this machine.
- EIGHT CHANNEL LPDDR5X - LPDDR5X is a new ground breaking memory small form factor installed on-board. With blazing speeds up to to 8000MT/s, it runs 1.5x faster than the DDR5 SODIMMs; 90% better performance over DDR5 SODIMMs in video conferencing and photo editing; 30% better performance in productivity apps; 12% better performance in digital content workloads.
- QUAD SCREEN 8K DISPLAY SUPPORT - EVO-X2 AI Mini PC support 4-screen 4K/8K output via HDMI 2.1 (8K@60Hz), DisplayPort 1.4 (4K@60Hz), and dual USB 4 40Gbps Transfer speed (supporting PD3.0/DP1.4/DATA). Ideal for gaming, video editing, and multitasking, it provides expansive and crisp multi-display support.
This is a contract problem, not evidence that the file appeared and disappeared. Markova summarizes the mechanism as “Ambiguity in, probability out.” A path is only deterministic when both producer and consumer agree on how to interpret it.
Why retries and run-level monitoring hid the problem
Retries can improve recovery, but they can also make an unreliable operation look healthy. In this account, a failed file read could be followed by another attempt that succeeded, while run reports recorded none of the failures. The traces remained in session transcripts. A final successful article therefore did not prove that every intermediate operation worked as intended.
Rank #2
- Built for Local AI Development: AMD Ryzen AI Halo is designed for local AI development and inference, featuring 128GB unified memory and support for up to 200B parameter models to build and run intensive AI workloads locally.
- 128GB Unified Memory: Features 128GB LPDDR5x unified memory at 8000 MT/s with 256 GB/s memory bandwidth, providing a shared memory pool across the CPU, GPU, and NPU to support larger AI models.
- AMD Ryzen AI Max+ 395 Processor: Features 16 cores, 32 threads, and Zen 5 architecture, paired with AMD Radeon 8060S integrated graphics featuring 40 RDNA 3.5 compute units and an AMD XDNA 2 NPU with up to 50 TOPS.
- Linux AI Developer Platform: Purpose-built for Linux-based AI development with full AMD ROCm software support and preloaded tools, models, and workflows optimized for local AI development.
- Compact, Connected Design: Includes a 2TB M.2 SSD, 10GbE LAN, Wi-Fi 7, Bluetooth 5.4, USB-C connectivity, and HDMI 2.1b.
For teams diagnosing agents, the distinction is between outcome monitoring and trace auditing:
- Outcome monitoring answers whether the run eventually completed.
- Transcript or trace review can reveal failed reads, retries, alternate-path guesses, and other recoveries that a success flag misses.
Neither the number of path attempts nor a per-attempt failure rate was established in Markova’s post. The transcript count should not be used to estimate either.
Rank #3
- EVOLUTION RYZEN AI MAX+ 395 MINI PC - GMKtec EVO-X2 is the next evolution in AI mini PC Ryzen Strix Halo series. Thanks to AMD Simultaneous Multithreading (SMT) the core-count is effectively doubled, to 32 threads. Ryzen AI Max+ 395 has 64 MB of L3 cache and can boost up to 5.1 GHz, depending on the workload. The Ryzen AI Max+ 395 is currently rated as the "most powerful x86 APU" on the market for AI computing.
- AI NPU with XDNA 2 ARCHITECTURE - Powered by 16 “Zen 5” CPU cores, 50+ peak AI TOPS XDNA 2 NPU and a truly massive integrated GPU driven by 40 AMD RDNA 3.5 CUs, the Ryzen AI MAX+ 395 is a transformative upgrade and delivers a significant performance boost over the competition. The Ryzen AI Max+ 395 excels in consumer AI workloads like the llama.cpp-powered application: LM Studio. Shaping up to be the must-have app for client LLM workloads, LM Studio allows users to locally run the latest language model without any technical knowledge required and unleash their creativity and productivity.
- AMD RADEON 8090S iGPU GAMING PC - The AMD Radeon RX 8060S offers all 40 CUs with up to 2.9 GHz graphics clock and uses the new RDNA 3.5 architecture. The powerful iGPU is positioned between an RTX 4060 and 4070 laptop GPU and therefore enables gaming in FHD at maximum details in most demanding games. The 8060S can also utilize the full 128GB pool, which is perfect for running LLMs such as Deepseek 70B Q8, which runs comfortably on this machine.
- EIGHT CHANNEL LPDDR5X - LPDDR5X is a new ground breaking memory small form factor installed on-board. With blazing speeds up to to 8000MT/s, it runs 1.5x faster than the DDR5 SODIMMs; 90% better performance over DDR5 SODIMMs in video conferencing and photo editing; 30% better performance in productivity apps; 12% better performance in digital content workloads.
- QUAD SCREEN 8K DISPLAY SUPPORT - EVO-X2 AI Mini PC support 4-screen 4K/8K output via HDMI 2.1 (8K@60Hz), DisplayPort 1.4 (4K@60Hz), and dual USB 4 40Gbps Transfer speed (supporting PD3.0/DP1.4/DATA). Ideal for gaming, video editing, and multitasking, it provides expansive and crisp multi-display support.
What else the audit reportedly found
Markova’s reported commit accounting describes other recurring workflow issues alongside the path errors. These are her figures, not independently verified measurements:
| Issue | Reported count | What the count means |
|---|---|---|
| Path-resolution failures | At least 117 | Lower bound from transcript review; not a rate per path attempt. |
| Writes to files never read first | 172 | Reported count of writes without a prior read. |
| Edits anchored to stale file content | About 20 | Approximate count reported by the author. |
| Sessions dispatching an LLM fixer for a formatting error a script could fix deterministically | About 19 | Approximate session count reported by the author. |
The adjacent examples point to a broader audit question: did the agent act on the intended resource and current information, or did the workflow merely recover well enough to finish?
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.How the reported fix changed both sides of the contract
Markova says the team changed the brief generator so it prefixed the worktree root and could no longer emit a relative path. They also updated all five agent contracts to say that paths should be used as provided and not joined with the working directory. In other words, the producer was made responsible for emitting an unambiguous value, and the consumer was instructed not to reinterpret it.
Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Scan for outdated or missing drivers - takes under a minute3Repair Windows errors before they cause bigger problemsThe author reports that a real generator verification produced 789 briefs with zero non-absolute paths, and that the TypeScript and Python test suites were green. Those results support the narrower claim that the generator was checked for absolute-path output; the post does not provide the briefs or code change for independent inspection.
What this incident suggests for agent workflows
The practical lesson is to put deterministic facts in deterministic code rather than relying only on a prompt to resolve ambiguity. For paths, decide the representation at the boundary and state how the consumer must treat it.
- Normalize at production: have the generator emit absolute paths if that is the contract, rather than asking the agent to infer a base directory.
- State the consumer rule: specify whether paths are absolute or relative and prohibit unrequested joining or rewriting.
- Audit recoveries, not just outcomes: inspect transcripts or traces for failed operations that retries may conceal.
- Verify the actual boundary: test generated briefs for the required format, then separately check that the consumer follows the contract.
These are design takeaways from Markova’s account, not a benchmark showing that one approach eliminates every agent failure. She says transcripts were not mined again after the change, so the post does not establish that the errors stopped appearing in later operation. Her closing formulation is that “Contracts for agents must be stricter than contracts for code.”
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
