That is awesome! Performance has definitely been my main issue when trying it out. To the point where I used /goal and asked to get startup on a small changeset below 150ms. In the end, the conclusion at the time was that OpenTUI was the main bottleneck.

GraphQLの実行に失敗した場合、CIで作成したキャッシュ（mob-sakai.github.io/unity-ch…）から取得してるので、5月下旬時点のデータは生きてる。 一旦、旧unity-changesetみたいにrss feedからデータベース更新するようにしないとなぁ...

NMP29 Liquibase Changeset!

Unityのリリース情報を得るために公式のGraphQL使ってるんだけど、5月下旬頃から404返すようになってる？ services.unity.com/graphql unity-changesetで使ってるから、勘弁してほしいなぁ...

Haha! I ran out of tokens in 10 minutes trying to do review of a 250 -40 python changeset. It's simply fabulous in generating cashflow for Anthropic.

I'm building a context engine for reviews and used @OpenAI Codex to its limits to stress test the graph and responsiveness of the commands on a large changeset whilst keeping the outcome relevant. The goal feature is super well though out, and with the context engine I'm building for the reviews and sourcing the relevant information for validation and correctness it managed to keep things on track. Now I need to wait a bit to reset my usage limits 😅

Here’s the updated thesis for $Xerg. @xerg_AI is building the FinOps layer for AI agents before anyone else realizes it’s needed. Every serious company running AI agents at scale has the same problem — they can see token counts but have no visibility into where dollars are actually leaking. Retry loops, bloated context windows, idle spend, and model overkill are draining budgets silently. Xerg turns that invisible waste into a dollar-denominated audit with one command. The GitHub is real. Pure TypeScript monorepo, Biome linter, Changeset versioning, Vitest, CI waste-rate gates. 98 commits, active releases, 3 contributors. This is not a demo project. Backed by a16z Scout, NVIDIA Inception, and Cloudflare Launchpad. Early institutional signal before a public raise. The core thesis: agent infrastructure is maturing fast and FinOps always follows compute adoption. It happened with AWS, it happened with Kubernetes, it will happen with AI agents. Xerg is first mover in the agent economic layer with a local-first, no-lock-in distribution model that removes all friction to adoption. Critically — Xerg already supports both OpenClaw and Hermes. This is not a single-runtime bet. Whichever agent framework wins the market, or if they split it, Xerg has parsers running on both. The economic audit layer sits above the runtime war entirely. Local-first free tier drives adoption. Hosted Pro converts teams that want shared history and CI integration. Clean bottom-up SaaS motion. Very early. Very low traction today. Very high upside if the agent infra thesis plays out.

I made some changes in my code and the tests stopped working. I asked an agent to make them work giving it the full changeset. It removed my changes in the code

Just telling my agent to changeset, typecheck, lint, git commit, push and spending $1 each time to make it happen

CopilotのReview改善、地味に嬉しい。 ・ コメントに High / Medium / Low の severity 分類を追加 ・ 類似コメントを自動グルーピング、""一度だけ指摘"" でノイズを削減 ・ Suggested changeset UI も刷新、優先順位付けが容易に github.blog/changelog/2026-0…

「Hunk」面向 Agent 编程的终端 Diff Viewer：先审查，再合并 它把多文件 changeset 放进可交互终端 UI，支持侧边栏导航、split/stack 布局、watch mode、Git/Jujutsu 和 stdin patch。 亮点是 inline AI / agent annotations，让 Agent 生成的改动更容易被人工复核。 👉 wefound.cc/p/2040.html

- I think I'd actually want to use voice to ask questions, and maybe get audio response back. I want the whole thing to feel like a really great developer is explaining me a huge changeset, layer by layer, with me being able to ask questions / go deeper whenever I want.

Useful technique in Amp with GPT-5.5 (deep mode) to make small independent commits from a big uncommitted changeset: "git stash the uncommitted (including untracked) changes, then reapply this refactor that you performed on HEAD, then git stash pop and fix merge conflicts"

Testing actual recommendations on workouts that can be applied automatically to following weeks. You'll get a changeset of the actual modifications if there are suggestions Following weeks happen to have a follow up as well