Windows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallCrashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minuteAn AI coding agent can start a new session without the conversation history from the last one. That is different from losing access to your repository, and different again from having project instructions available but failing to follow them. The practical fix is to make important, stable project facts available through supported repository instructions or memory features—while treating them as guidance, not a guarantee of recall or correctness.
The title’s “what I built” does not identify a tool, implementation, compatibility list, or test result. Without those details, there is no basis to describe or evaluate a specific build. The underlying problem, however, has several concrete causes and ways to address them.
As an Amazon Associate I earn from qualifying purchases.
What “forgetting the project” can mean
People use “forgetting” to describe different failures. Identifying which one is happening helps you choose a fix rather than adding more notes blindly.
Do these 3 things before closing this tab:
1Scan for outdated or missing drivers - takes under a minute2Repair Windows errors before they cause bigger problems3Fix the driver behind crashes, sound loss and screen glitches- Conversation loss: A fresh session does not include the prior session’s messages. Claude Code documentation says, “Each Claude Code session begins with a fresh context window.” That is a statement about the session context, not proof that the agent cannot read the same repository again. Claude Code: How Claude remembers your project
- Missing project context: Relevant facts—such as architecture, conventions, or test commands—were never stored in a file or memory feature that the new session loads. Copilot CLI describes its context window as the material available to the model for a response, including conversation messages and tool activity. GitHub Docs: Managing context in GitHub Copilot CLI
- Instructions not followed: The agent may have access to the right guidance but interpret or apply it inconsistently. GitHub cautions that Copilot may not follow custom instructions exactly the same way every time. GitHub Docs: About customizing GitHub Copilot responses
These cases can overlap. A project file can carry durable facts into a new session, but it cannot recreate every detail of an old conversation or ensure that every instruction affects every answer.
#1 Best Overall
How to keep useful project context between sessions
Put stable, repeatedly needed facts in version-controlled repository instructions. A concise file can tell an agent what the project does, where important components live, which conventions matter, and which commands or workflows to use. GitHub says a maintained AGENTS.md or .github/copilot-instructions.md can give agents a structural overview so they need not read large numbers of files just to orient themselves. GitHub Docs: Optimizing your AI usage to maximize efficiency and reduce cost
Keep this guidance focused on information that remains useful across tasks. Recent task state—what you tried, what failed, and what remains—changes more often, so record it separately in a short handoff or session summary when needed. Keeping stable project facts apart from temporary task notes makes it easier to update each without turning one file into a stale transcript.
GitHub documents both repository-wide and path-specific instructions. Path-specific guidance can keep specialized rules from being applied to unrelated parts of a codebase. GitHub Docs: About customizing GitHub Copilot responses
Does AGENTS.md work across coding agents?
AGENTS.md is an emerging interoperability convention, not a universal guarantee that every agent will discover and apply the file. Loading behavior depends on the tool, its version, configuration, and instruction precedence. Claude Code documentation describes version requirements and settings that affect AGENTS.md loading, as well as its interaction with CLAUDE.md. Check the current documentation for the agent you use rather than assuming a filename alone is enough. Claude Code: How Claude remembers your project
Rank #3
An exploratory 2026 study of 2,926 GitHub repositories reports that context files were the dominant configuration mechanism among the tools it studied, and that AGENTS.md was emerging as an interoperable standard. The same study describes shallow adoption of skills and subagents in its sample. This is evidence about those repositories and tools, not proof of universal support or consistent loading. Harness Engineering for Agentic AI Coding Tools: An Exploratory Study
Why an agent may ignore instructions
First verify that the file is in a location the tool recognizes and that the tool version and settings support loading it. Then check whether the instruction applies to the files or task at hand, and whether another instruction source takes precedence. If the expected guidance is loaded and applicable, inconsistent adherence may still occur; it is not necessarily evidence that the file failed to load.
Keep instructions specific and actionable. “Use the project’s existing test command before proposing a change” is more useful than a vague request to be careful. Avoid duplicating conflicting rules across global, repository-wide, and path-specific files. GitHub’s own warning is worth taking literally: “Due to the non-deterministic nature of AI, Copilot may not always follow your custom instructions in exactly the same way every time they are used.” GitHub Docs: About customizing GitHub Copilot responses
What the evidence says about whether context files help
Context files can reduce repeated orientation work by making relevant facts available, but availability is not the same as improved code correctness. A 2026 study, Do Context Files Help Coding Agents? A Two-Agent Ablation Study on Real Repositories, evaluated 288 runs across 17 real tasks and three repositories. For the two agents tested—Claude Code and Codex—the authors found no measurable correctness movement from context strategy under their equivalence test, which bounded effects to no more than 10–15 percentage points. This is a bounded result for that study’s agents, tasks, repositories, and method; it does not establish that context files never help or that other outcomes, such as reduced setup effort, cannot change. Do Context Files Help Coding Agents? A Two-Agent Ablation Study on Real Repositories
No broad rate of coding-agent “forgetting,” measured productivity savings, or performance improvement for the unspecified build in the title is established by these sources. Claims about a particular implementation would require its documentation and evidence.
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.
