You can also use codex in the terminal with a persistence context, you start your first session and make it try to learn and understand the project context, draft an md file from what it learns, review and refine, repeat and update the file - for example AI_CONTEXT.md- until you have a solid starting context, add an instruction in the file to update it with new learnings an findings from every subsequencce sessions, add a shell script what ever you like in your project to load codex with that file as context, from that point on you will have an incrementally built persisted context and you dont worry about the agent forgetting all the context you had before., this might also work with any agent as long as you can invoke it with a pre-defined context.
I found this useful for my case. On Sunday, January 4, 2026 at 4:20:25 AM UTC+3 Craig Mitchell wrote: > If you use IntelliJ, I highly recommend the Windsurf plugin. It allows > you to use any model you like. ChatGPT, Gemini, Claude Sonnet, ... > > And as Jens says, it reads your project, so it knows what you're doing, > and it's scarily good at helping you out (refactoring, or doing anything). > > Privacy and keeping intellectual property private probably goes out the > window though. If you need that, you'd need a sandboxed Copilot > subscription, or something similar. > > On Sunday, 4 January 2026 at 4:20:53 am UTC+11 Tim Macpherson wrote: > >> I asked chatgpt about GitHub copilot and Gemini CLI, Re complex >> refactor where I'm stripping out pure java logic from a GWT client module >> incrementally in a correct order so the app isn't broken at any stage: >> Chatgpt also says: your instinct — “maybe AI can’t do that” — is correct. >> It uses an analogy: >> AI is good at: rebinding books, rewriting titles, rearranging whole >> shelves >> It is bad at: Knowing which single book is safe to slide now. >> That’s human judgement grounded in system knowledge. >> >> Yahoo Mail: Search, Organize, Conquer >> <https://mail.onelink.me/107872968?pid=nativeplacement&c=US_Acquisition_YMktg_315_SearchOrgConquer_EmailSignature&af_sub1=Acquisition&af_sub2=US_YMktg&af_sub3=&af_sub4=100002039&af_sub5=C01_Email_Static_&af_ios_store_cpp=0c38e4b0-a27e-40f9-a211-f4e2de32ab91&af_android_url=https://play.google.com/store/apps/details?id=com.yahoo.mobile.client.android.mail&listing=search_organize_conquer> >> >> On Sat, Jan 3, 2026 at 4:13 PM, Jens >> <[email protected]> wrote: >> >> How do you use Gemini / ChatGPT? Via some IDE plugin or just their >> website? >> >> For example if you use Github Copilot and use the official IDE plugin >> then you can store markdown files in your project which describe your >> stack, coding patterns, etc. The plugin will then read these files and add >> the content to the context of each chat session you start. That way the >> model always knows these information. These files are different for models >> / plugins, e.g. Gemini CLI searches for GEMINI.md files if I remember >> correctly. You have to read the docs of the tools you are using. >> >> These language models are good at pattern recognition and pattern >> completion. They cannot read your mind nor have a mind. What matters is >> that the context passed to the model is good and that it can see clear >> instructions so it must not guess. Only then the output will be good. >> >> -- J. >> >> Tim Macpherson schrieb am Freitag, 2. Januar 2026 um 19:30:31 UTC+1: >> >> Gemini just output that for a complex refactor where I'm stripping out >> pure java logic from a GWT client module incrementally in a correct order >> so the app isn't broken at any stage: AI can't do that, agents or >> whatever. Maybe in 5 years, I have no idea >> >> Yahoo Mail: Search, Organize, Conquer >> <https://mail.onelink.me/107872968?pid=nativeplacement&c=US_Acquisition_YMktg_315_SearchOrgConquer_EmailSignature&af_sub1=Acquisition&af_sub2=US_YMktg&af_sub3=&af_sub4=100002039&af_sub5=C01_Email_Static_&af_ios_store_cpp=0c38e4b0-a27e-40f9-a211-f4e2de32ab91&af_android_url=https://play.google.com/store/apps/details?id=com.yahoo.mobile.client.android.mail&listing=search_organize_conquer> >> >> On Wed, Dec 31, 2025 at 4:48 PM, 'Tim Macpherson' via GWT Users >> <[email protected]> wrote: >> >> Thanks, I heard the agents in cloud can be a security risk for >> proprietary code, but there are options to keep it private. Thinking about >> it an agent would be overkill for my purposes, I just want something that >> doesn't have to be constantly corrected because it's forgotten my current >> stack. It can just about remember that I'm using Java GWT and D3 js, memory >> of a 4 year old >> >> Yahoo Mail: Search, Organize, Conquer >> <https://mail.onelink.me/107872968?pid=nativeplacement&c=US_Acquisition_YMktg_315_SearchOrgConquer_EmailSignature&af_sub1=Acquisition&af_sub2=US_YMktg&af_sub3=&af_sub4=100002039&af_sub5=C01_Email_Static_&af_ios_store_cpp=0c38e4b0-a27e-40f9-a211-f4e2de32ab91&af_android_url=https://play.google.com/store/apps/details?id=com.yahoo.mobile.client.android.mail&listing=search_organize_conquer> >> >> On Wed, Dec 31, 2025 at 3:54 PM, Jens >> <[email protected]> wrote: >> >> What is your goal? AI agents running locally are usually a security >> nightmare (data deletion, secret exfiltration via prompt injection) so you >> have two choices: >> >> 1.) babysit them all the time and review, understand and confirm any >> command the agent wants to execute. Otherwise it might delete your drive >> accidentally and feels really sorry about it or it reads an untrusted >> website with some prompt injection sections which causes the model to do >> something totally different. As a middleground you can allow read/search >> commands by default but still review everything else. >> >> 2.) Run the agent autonomously in a secure environment to protect your >> local data. This can be a local VM, a local docker container (e.g. use >> devcontainers for development in general or just run the agent in a >> container) or an OS provided sandbox mechanism. Then you make only the >> folders accessible to the environment (and thus the AI agent) which are >> required. If your agent should use the internet for research then you must >> control network requests (e.g. allowed domains). >> >> Depending on the code size and the task complexity an AI agent is >> relatively slow as it does a lot of research in the code base first before >> implementing stuff. So if you work on a laptop it might be annoying to wait >> for the agent to complete. So the last option is using agents in the cloud >> which also solves the security issue for your local data as the cloud >> solutions usually must clone your repository from somewhere. >> >> -- J. >> >> Tim Macpherson schrieb am Dienstag, 30. Dezember 2025 um 20:25:14 UTC+1: >> >> >> >> Following on from a recent post about AI, I use basic chatgpt or >> Gemini. Should I be using AI agents such as Claude Code, OpenAI’s Codex, >> etc ? My current AI usage is ok for specifics but awful at remembering >> things. >> >> Yahoo Mail: Search, Organize, Conquer >> <https://mail.onelink.me/107872968?pid=nativeplacement&c=US_Acquisition_YMktg_315_SearchOrgConquer_EmailSignature&af_sub1=Acquisition&af_sub2=US_YMktg&af_sub3=&af_sub4=100002039&af_sub5=C01_Email_Static_&af_ios_store_cpp=0c38e4b0-a27e-40f9-a211-f4e2de32ab91&af_android_url=https://play.google.com/store/apps/details?id=com.yahoo.mobile.client.android.mail&listing=search_organize_conquer> >> >> -- >> You received this message because you are subscribed to the Google Groups >> "GWT Users" group. >> To unsubscribe from this group and stop receiving emails from it, send an >> email to [email protected]. >> To view this discussion visit >> https://groups.google.com/d/msgid/google-web-toolkit/9585ff0f-015e-4344-931f-6610ae632644n%40googlegroups.com >> >> <https://groups.google.com/d/msgid/google-web-toolkit/9585ff0f-015e-4344-931f-6610ae632644n%40googlegroups.com?utm_medium=email&utm_source=footer> >> . >> >> -- >> You received this message because you are subscribed to the Google Groups >> "GWT Users" group. >> To unsubscribe from this group and stop receiving emails from it, send an >> email to [email protected]. >> >> To view this discussion visit >> https://groups.google.com/d/msgid/google-web-toolkit/546002399.7992174.1767199698829%40mail.yahoo.com >> >> <https://groups.google.com/d/msgid/google-web-toolkit/546002399.7992174.1767199698829%40mail.yahoo.com?utm_medium=email&utm_source=footer> >> . >> >> -- >> You received this message because you are subscribed to the Google Groups >> "GWT Users" group. >> To unsubscribe from this group and stop receiving emails from it, send an >> email to [email protected]. >> To view this discussion visit >> >> >> https://groups.google.com/d/msgid/google-web-toolkit/a2cad606-9195-4a27-8e11-3ca5119f6748n%40googlegroups.com >> >> <https://groups.google.com/d/msgid/google-web-toolkit/a2cad606-9195-4a27-8e11-3ca5119f6748n%40googlegroups.com?utm_medium=email&utm_source=footer> >> . >> >> -- You received this message because you are subscribed to the Google Groups "GWT Users" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To view this discussion visit https://groups.google.com/d/msgid/google-web-toolkit/9bc1bbbf-8e74-4c80-a8ce-e57d4759b96fn%40googlegroups.com.
