Activity
Mon
Wed
Fri
Sun
Jun
Jul
Aug
Sep
Oct
Nov
Dec
Jan
Feb
Mar
Apr
What is this?
Less
More
Memberships
Clief Notes
26.2k members • Free
15 contributions to Clief Notes
13d •
Finally got to rewriting poor man’s memory for opencode
It’s not perfect, and I appreciate those of you who tried out the original PMM-plugin (would appreciate the feedback from that). TLDR (AI summarised): Shipped the rewritten Poor Man’s Memory harness — now works across both Claude Code and opencode (Kimi K2.5, Gemini Pro/Flash) from the same project. Choose the model that fits the task, keep the memory. Tested in a multi-user Telegram channel holding concurrent 1:1 threaded conversations without cross-contamination. Previously validated switching between codebases mid-session. Link: github.com/NominexHQ/pmm-harness-dist Looking for early feedback and bug reports. Also built an agentic skill-porting suite between Claude Code and opencode as the first serious test of the rewrite. — The original plugin was Claude code only and worked well for me when I needed to switch beteeen code, CLI and Cowork (and between models) while retaining its memory. I tested it in a multi user environment (a telegram channel) and it held up holding multiple 1:1 individual threaded discussions with users a single group without straying. I had already been using it to switch between different codebases in a single session. So I had the suspicion it would hold up in such noisy environments. I just could test it until I implemented the telegram integration. This next one allows me to run my memory in opencode (where I run kimi-k2.5, Gemini pro and Gemini Flash) and Claude code from the same project. I get to choose which models or apps untike suits what I’m working on at any given moment. Would appreciate some early feedback and bug reports. I’ve been occupied rewriting Vera for opencode and realised I needed a plugin that allows me to analyse and port skills and plugins between the two. Writing the agentic suite for doing this was my first serious test of the rewritten PMM implementation.
1 like • 6d
@Millenial Cat What you're describing is really interesting. I plan on reading your research paper to get a better understanding, but its always subject to time. Definitely procedural. I'm a tax accountant, I need to solve problems for people. Primary concern for the legislative side of things is how chunk the data. Killing the context by cutting a section in half won't do. The structured nature of the source material should make that workable though. I'm planning a layered approach, Legislation->Case Law->Tax office rulings & guidance->Professional commentary. From authoritative to interpretative, with weighting built into the search results. The goal is to have the kind of assistant function you mention. I don't need a database that can just find a particular section, I can just go look that up. Correct me if I don't understand, but you're saying that PMM isn't an archival system itself? Could it form part of the assembly process for the knowledge layer? Right now this part of my project is all ideas while I work on some simpler things and build my skills and knowledge in these systems.
1 like • 6h
@Millenial Cat Australia. Common law system like you, so, yes, court decisions set precedent, with appeals to higher courts. We have a federal system though, so each state has its own parliament and courts. Pretty much all the income tax laws are federal, so I rarely have to deal with state law. So, legislation and case law are the authority. The tax office administers and enforces the law, so their rulings and guidance are generally considered 'the law', although they are always defending their positions in the courts. This I think might actually be the most difficult part of this project. The tax law landscape is like quicksand. No idea how much work it might be keeping things up to date from a data ingestion perspective. Either way, I think PMM might be interesting even for managing my day to day for what I'm doing and need to track for client work. I really like your post about the update to PMM. I'm going to need to look at it some more. I'm not sure about some other 'memory' structures. Congnee seems overkill. Nate B Jones's open brain might be too. Plus I think it uses postgres, which wasn't in the plan.
6d •
The Person Using The Tool Should Be The One Making It
Most software is built by people who will never use it. They are guessing what their users need, then guessing whether they got it right. That gap is why so many tools feel almost-right. Almost the workflow you wanted. Almost the shortcut you would have built yourself. The fingerprint is missing because the maker's hands were never in the work. Pushing Squares started from the opposite direction. I am a colourist and a director. CRUSH, CBA, FUZZ, SUBTRACKT, REFRAKT. These plugins exist because I needed them on a real timeline that day. The first user is always me. The shipping bar is "would I install this tomorrow on a paying job". This changes the design surface in three ways: 1. The defaults are right because the maker lived inside them. Setting a default is a guess about the most common case. If you are the most common case, you are not guessing. 2. Features earn their slot in the panel. Anything that does not save time on a real shot gets cut. There is no marketing reason to keep a feature the maker themselves never reaches for. 3. The bug list is short and honest. You feel every regression because you hit it the next morning at 2am. This is not a no-code argument. It is not "everyone should learn to ship code". It is a positioning argument. The closer the maker is to the work, the less translation has to happen between intention and tool. AI changes the cost of this. Building the thing you needed used to take a development team and six months. Now it takes a directed run with Claude and an afternoon. The bottleneck moves from technical capability to taste, judgment, and knowing what the work actually demands. So if you are using a tool every day and feeling the gap, the gap is the brief. The next person to fix it should be you. The closest hand to the work makes the truest tool_ // A<3
1 like • 6d
@Ari Evergreen Time will tell I guess. They're talking rollout at the end of June. I'll know by then if it doesn't have another delay. They've hired a consulting firm and this is their business, so who am I to argue right now. The Microsoft ecosystem isn't so bad. Its what we had at my last job. I know some are fussy about that stuff, but it did everything I needed it to. Also, sharepoint etc. tends to have better integrations with the other tools and tech in our industry. Apparently.
1 like • 7h
@Ari Evergreen Small update? They dropped the change to outlook on Wednesday. No big thing, just email, easy switch for everyone. But then yesterday they ran a 1 hour session on the new DMS/CRM from 12 noon to 1pm. An email was sent just before it saying that the old file system, which was just a network drive, would become read only at 3pm. This was the first actual info other than the name anyone got, I think. Yesterday after the session, I was talking with someone in the lunch room and I said that people are going to simply avoid anything that gives them the littlest bit of friction and will simply work around them. Today I began hearing about a few cases of exactly this happening. https://fyi.app/ if you're at all interested. There's a few more things to come which will roll out between here and July. I didn't expect anything just yet.
19h •
8 Hours, 5 Sessions, One Site
So, after people's comments about longer breakdowns and more graphics in my posts, I decided to make a site where I can share downloads and more visual breakdowns of work I'm doing. 8 hours from concept to live site. 5 sessions. None of them above 50% context. The site is https://www.aris-space.com/ Built on Next.js 15, Tailwind v4, MDX, Zustand and react-rnd, deployed on Vercel. Full breakdown here VVV https://www.aris-space.com/documents/discipline-and-process/building-aris-space I architected. Claude executed. // A<3
1 like • 8h
I love the site design. Takes me way back to mac and 3.11 days. Full disclosure, once the focus on the functionality for the little dashboard shell + tools I'm building are satisfactory, I'm totally stealing this idea. I'm no developer, but mine's just a shell served up by a flask server locally, so hopefully Claude and I can make it work. There's so much to learn.
Mar 12 •
Folder Structure
Did anyone try folder structure yet?What is the most difficult step in this architecture?
0 likes • 6d
@Aaron Quiroz I feel you. My 4 are still in the 12yrs down to 5yrs range.
0 likes • 6d
@Vincent van Beek When I first found Jake's YT channel and then Clief Notes, I just took the URL for both and threw it into the browser claude chat where I'd been brainstorming my idea and asked it what it thought, and when we'd worked a few specifics I had it create me the Layer-0 and Layer-1 files, along with a conventions.md which I have it refer to when we set up something new. Its initial job was pretty good, although I plan to tighten up the conventions with proper reference files so that we don't get too much drift.
11d •
your AI knows exactly where your understanding breaks down.
Feynman said if you can't explain it simply, you don't understand it. i took that literally and started explaining concepts to claude like it's never heard of them. no jargon, no shortcuts. plain language only. what i didn't expect.. claude pushes back. and it pushes back exactly where the explanation gets fuzzy. that's not a coincidence. that's the gap. Been running this on jake's material. lesson lands, i explain it back, claude presses on the parts that don't hold. what i thought i understood and what i can actually defend out loud are two very different things. The AI isn't teaching you. it's exposing what you never actually learned. anyone else using AI as a sparring partner?
1 like • 6d
@Samuel Albert Love the brainstorming mode idea set up as an actual workflow. Definitely going in the todo list. Forget the AI, maybe it can be a guardrail for me to not be lazy. I had the most surreal conversation with Claude Code the other day. It wasn't brainstorming, but serves an important reminder. I had asked claude chat about opusplan mode and it gave a good simple answer and explained how to use it. When I got home I couldn't find how to do it in Claude code, so I asked CC how to access it and had the following conversation (paraphrased): Me: I can find the options to set opusplan mode, where is it? CC: Oh yeah, just /model and select it from the list. Me: It's not in the list, and I can't find mention of it anywhere in settings. CC: Oh, sorry, I've checked and it doesn't exist. You can only switch the models manually. Me: I had this conversation earlier today. (Paste the earlier chat). CC: Oh, I'm sorry, I was wrong. There must be a bug in the interface, we can just set it as default in the settings.json file, then it should work. Me: Yeah, still not working. CC: You know, I totally think that other conversation was hallucinating. Just change the models manually when you are finished in plan mode. Me: Ok, what about this info from Anthropic? CC: Oh, let me check. Oh, yeah its totally real, but its been removed. Here's a support ticket requesting to bring it back....etc,etc,etc. Me: You know what, that's 45mins of my life wasted. I'm going to bed now. Lesson learned, again. Don't be like me. Don't let AI make you chase ghosts. Context first and always.
1-10 of 15
@jarad-taylor-7582
Father of 4 and accountant, looking to redefine the way I do my work with the most simple and capable tools I can find.
Online now
Joined Apr 15, 2026
Powered by