Nima Aksoy

این موسیقی که باهاتون شر کردم رو مدل Vocaler v0 که داریم آموزشش میدیم ساخته. اینجا اولین جاییه که اشتراک گذاشتم. ویدیو رو با سیدنس ۲.۵ ساختم. داخل وکالر همه مدلهای مهم ویدیو و عکس هم اضافه میشه که کاربرها بتونن ویدیو و کاور تصویر بسازند

دوستان این چند وقت ساکتم دارم رو پروژه vocaler کار میکنم که به زودی لانچ میشه تو وکالر می‌تونید به همه مدل های موسیقی دسترسی داشته باشد، پروفایل خودتون رو بسازید و موسقیهاتون رو پابلیش کنید غیر از اون ما داریم مدل ساخت موسیقی خودمون رو هم درست میکنیم تا الان ۵۰۰ نفر ثبت نام کردن و نام کاربریشون رو ساختن از اینجا می‌تونید ثبت نام کنید و دوستاتون رو با لینک اینوایت خودتون دعوت کنید هرکی که دعوت می‌کنید به فالورهای شما هم اضافه میشه vocaler.ai/nima/invite

cost-xray is the missing layer under Claude Code and Codex bills. It captures the real API request and shows which schema, MCP, or tool call owns the tokens — not just a session total from logs. https://nimaaksoy.com/radar/cost-xray

Treg is OpenRouter for agent tools. One token, thousands of endpoints, pay per call, and your own keys stay on the server while every teammate agent can use them. https://nimaaksoy.com/radar/treg

Google open-sourced AX — kubectl for agent workloads. YAML tasks, sandboxed runs, workspace + MCP wiring, and egress fences so multi-agent fleets stop looking like ungoverned chat loops. https://nimaaksoy.com/radar/ax

Univer is the office harness agents actually needed. Sheets, docs, slides in one SDK — same Facade API in the browser and headless on Node, with CLI/harness plugins so people review agent drafts in the same files.\n\n https://nimaaksoy.com/radar/univer

Laya is a real System 1 layer for agents. Typed choice/score/yes-no over any ticket or JSON in ~33 ms — no free-form text, nothing to parse, multilingual router included. https://nimaaksoy.com/radar/laya

json-render is the Generative UI framework that finally feels production-safe. AI only emits JSON against your component catalog — streamable screens, no freestyle JSX mess. Vercel Labs, open source. https://nimaaksoy.com/radar/json-render

OpenShell is the sandbox layer coding agents have been missing. NVIDIA’s open runtime puts Claude Code, Codex, and friends behind YAML policies for files, network, and keys — so the agent works without owning your machine. https://nimaaksoy.com/radar/openshell

Skill packs are the new soft underbelly. NVIDIA’s SkillSpector scans agent skills and MCP installs for prompt injection, exfil, and supply-chain junk before they land in Claude Code or Codex. https://nimaaksoy.com/radar/skillspector

Skill packs are the new soft underbelly. NVIDIA’s SkillSpector scans agent skills and MCP installs for prompt injection, exfil, and supply-chain junk before they land in Claude Code or Codex. https://nimaaksoy.com/radar/skillspector

Skill packs are the new soft underbelly. NVIDIA’s SkillSpector scans agent skills and MCP installs for prompt injection, exfil, and supply-chain junk before they land in Claude Code or Codex. https://nimaaksoy.com/radar/skillspector

Claude-Mem is the missing memory layer for coding agents. it captures what the agent did, compresses it, and puts the right context back into the next session on Claude Code, Codex, Hermes, and more. https://nimaaksoy.com/radar/claude-mem

OpenResearch is a local workspace that turns Claude Code, Codex, OpenCode, or Cursor into real research agents. Parallel directions, isolated worktrees, experiment archives — research that behaves like a repo, not one endless chat. https://nimaaksoy.com/radar/openresearch

OpenResearch is a local workspace that turns Claude Code, Codex, OpenCode, or Cursor into real research agents. Parallel directions, isolated worktrees, experiment archives — research that behaves like a repo, not one endless chat.\n\n https://nimaaksoy.com/radar/openresearch

OpenResearch is a local workspace that turns Claude Code, Codex, OpenCode, or Cursor into real research agents. Parallel directions, isolated worktrees, experiment archives — research that behaves like a repo, not one endless chat.\n\n https://nimaaksoy.com/radar/openresearch

RTK is a small Rust binary that filters bash output before your coding agent reads it. git, tests, ls, docker — less noise in context, lower token burn, no new model. https://nimaaksoy.com/radar/rtk

shadcn/lint is the missing hard check for agent UI. Tailwind design-system rules that Oxlint/ESLint can fail, with errors that tell the model which size or theme token to use instead of random p-4. https://nimaaksoy.com/radar/shadcn-lint

Open Code Review is Alibaba’s real PR review stack, now open source. Deterministic file pick + bundling + line placement, then an LLM agent for judgment. Higher precision, ~1/9 the tokens vs a generic coding agent on their bench. https://nimaaksoy.com/radar/open-code-review

Worktrunk is the missing layer when you run agents in parallel. Git worktrees addressed like branches, one-liners to spin Claude/Codex per feature, hooks, merge, and cleanup without path hell. https://nimaaksoy.com/radar/worktrunk