... noticed a different kind of context problem while working with AI coding agents. The agent can read the code. It can inspect the githistory. It can find the files it needs. But sometimes it still misses something important: Why we made a particular decision in the first place. For example: Why we deliberately didn't use an abstraction Why an API is handled differently in one workflow Why a certain dependency is pinned Why a seemingly duplicated piece ...

... with a great median and a fat tail every time on real work. Stephanie's time-to-merge is the same instinct measured in hours instead of tokens, with the advantage that it is already sitting in your githistory. Comment from Stephanie Tang(@stephanie_t0317): Token efficiency is an interesting metric, but I think the real metric is cost per outcome. A tool that uses fewer tokens but requires multiple rounds of debugging may be more expensive in terms ...

... team has been spending a lot of time thinking about existing products rather than brand-new ones. One thing we've noticed is that as a product grows, knowledge becomes fragmented. Some decisions live in documentation, some in Githistory, and some only in the minds of the people who built the feature months ago. Adding a "small feature" often means spending more time understanding the existing system than actually implementing the change. That's exactly the problem we wanted ...

... build things? In Claude, some other harness? Comment from Mustafa Arian(@thenameisarian): if i got unlimited Fable for a month, i'd build the thing i've been backlog-shelfing for 18 months: a private agent that watches my githistory, slack threads, and PR review comments, and writes the receipts of what i actually did this quarter for performance reviews + investor updates. half of an engineer's real work never lands on a resume because nobody writes it down ...

... this is compact contract identity, not full assumption-level versioning yet. So a receipt can anchor the run to the contract in play, but for the exact assumption that changed you would still pair it with contract diff and Githistory. That is also the direction we are pushing: stronger receipt evidence, without pretending the receipt alone is already the whole change ledger. Curious what you'd want the ideal artifact to look like there: a receipt that carries ... ... contract snapshot used, assumption that changed, command evidence, and whether the fix updated the contract or only local state. That s the drift I d pressure-test. Comment from Ahmad(@blah_mad): Today it s still a stitched artifact:

... forward." You can turn it on when the need becomes clear. You lose some history, but the tooling from that point forward is complete. AI code provenance does not work this way. The prompt a developer submits to Claude Code exists for a few hundred milliseconds in transit. After the model returns its response and the editor applies the change, that prompt is gone. Git

... good. So the one load-bearing question back, because it decides whether the code signal is even reachable for you: does NEST sit close enough to the repo to ingest githistory, or is it deliberately architected upstream of the code layer? If you can reach

... local repository is rarely just source code. Over time, it becomes a mix of source files, Githistory, dependencies, generated build output, caches, logs, assets, local databases, framework artifacts, and tool-specific folders. That makes project size surprisingly hard to reason about. When a repo grows to 10 GB, 30 GB, or more, the first question should not be: What can I delete? It should be: What is this project actually made of? That idea became Projects in Room Service ... ... Projects gives local development folders their own workspace. It scans repositories and breaks them down into meaningful groups like

... widely seen as the best. And Sonnet is definitely a high-quality model. BUT if you re building a SaaS product powered by AI, pricing is important. It directly impacts margins, scalability, and how competitive you can be. If alternatives are only a few percent behind at 3 7x lower cost (+ you can train them to achieve better quality), that gap is hard to ignore. If there is a way to bring Sonnet closer to Haiku pricing, that ... ... most tasks. - Frontend/UI work? Gemini Pro surprised me with speed and quality. The real game-changer isn't just the model though - it's how you structure your prompts and manage context. I've found that keeping a clean git

... code patterns that makes the code highly coupled, the way to refactor them and advices on the paradigm to use (functional, imperative, declarative...), identify the communication protocols between layers and all the dependencies, internals and on the frameworks - a githistory-based analysis to explain the legacy code. In the long term, it could replace $50k+ consulting engagements with a tool. I wonder if it worth building it and spend time on this. For those in software companies: - Have ...