Loading video...

Video Failed to Load

There was a problem loading this video. This could be due to a temporary network issue or the video might be unavailable.

Introducing InsForge Backend Branching. We made agentic coding production-safe. Your agent creates isolated backend branches when working. Every backend service, policy, and config is copied into a secure, isolated environment. Review PR-style diffs. Merge only when ready.

InsForge

2,795 subscribers

62,391 views • 4 months ago •via X (Twitter)

Education Science & Technology

Anya Rossi• Live Now

Private livecam show

31 Comments

Hitesh Choudhary4 months ago

Do you guys even sleep or just keep pushing new features? 😂

InsForge4 months ago

we'll catch up after launch week is done!

InsForge4 months ago

Try Branching: Read more:

Hang Huang4 months ago

We’ve been working on this for a long time This has meaningfully improved our own developer experience building with coding agents among our internal team, excited to see what you guys think!

InsForge4 months ago

can't imagine building without it now!

Henry4 months ago

Amazing!

ahmed abid4 months ago

gimme a break , i can't keep up T_T too much updates xd

InsForge4 months ago

hahaha keep up or get left behind!

crypticsaiyan4 months ago

Agentic coding we can actually trust. Huge! 🔥

Tony Chang4 months ago

safely allow agents operate!

InsForge4 months ago

yes sir!

dale4 months ago

Lfg!!! How much sleep did yall miss to ship this many stuff

InsForge4 months ago

not too too much...

Artists Voyage 🔶4 months ago

This is a smart way to remove the fear from agent driven changes.

InsForge4 months ago

try it and let us know what you think!

Matt4 months ago

isolating the backend per agent run is the right instinct. the thing that still bites is the agent's own context surviving across those branches, the work stays safe but the session state usually doesn't.

InsForge4 months ago

we copy everything

Matt4 months ago

copying everything also copies the stale state, that's the trap written with ai

InsForge4 months ago

we use 3 way state diff to manage read our blog! we're pretty open about everything

Matt4 months ago

3-way diff handles the data side cleanly. the part that slips is the agent's working memory: it merges the backend fine but keeps planning against the branch it remembers, not the merged result. written with ai

Gregor4 months ago

curious what 'copied' means for stateful services. does the DB snapshot at branch time, or does the agent still touch live data?

Tornike Sirbiladze4 months ago

branching for backend state is exactly the kind of boring infra agents need otherwise “agent changed config” becomes terrifying very fast

InsForge4 months ago

we are on our safety and reliability arc right now 😎

Taro Fukuyama4 months ago

Let's gooo!

Sridhar A4 months ago

isolated backend environments are an excellent way to make agentic coding safe.

InsForge4 months ago

ship 100x faster and safer

Alice The Ai Expert4 months ago

InsForge makes agentic coding safe with isolated backend branches.

Kyle4 months ago

Replace supabase?

InsForge4 months ago

think more like replacing AWS 👀 we have all the primitives like db, auth, storage, and so much more we handle your frontend deployments, provision any compute you need, LLM model gateway, stripe integration for payments, posthog integration for analytics, all native for agents

SuperPublic4 months ago

Backend Branching is such a useful unlock for agentic coding — congrats on shipping it!

InsForge4 months ago

thank you! it's been a game changer for us internally as well

Related Videos

Introducing Diagnose and Debug. Give your agent first-party backend context: health metrics, logs, errors, database checks, and more. Your coding agent catches backend issues now, then fixes them by working with your dedicated InsForge debugging agent. Agent <> Agent.

Introducing Diagnose and Debug. Give your agent first-party backend context: health metrics, logs, errors, database checks, and more. Your coding agent catches backend issues now, then fixes them by working with your dedicated InsForge debugging agent. Agent <> Agent.

InsForge

53,090 views • 4 months ago

Introducing InsForge Compute. Deploy long-running backend services as Docker containers via your coding agent. It writes the Dockerfile, builds the image, injects env vars, runs the container, and returns a live URL. All in under 60 seconds.

Introducing InsForge Compute. Deploy long-running backend services as Docker containers via your coding agent. It writes the Dockerfile, builds the image, injects env vars, runs the container, and returns a live URL. All in under 60 seconds.

InsForge

1,511,264 views • 4 months ago

.InsForge turns your coding agent into a principal backend engineer. Backend servers, database, LLM gateway, frontend deployment, and more. Every primitive built for how coding agents actually work, so they can run your entire infrastructure. Congrats on the launch, Hang Huang ☁️ & Tony Chang ☁️!

.InsForge turns your coding agent into a principal backend engineer. Backend servers, database, LLM gateway, frontend deployment, and more. Every primitive built for how coding agents actually work, so they can run your entire infrastructure. Congrats on the launch, Hang Huang ☁️ & Tony Chang ☁️!

Y Combinator

152,008 views • 5 months ago

Introducing InsForge 2.0: The Backend for Agentic Development Our OSS backend provides databases, auth, storage, model gateway, and edge functions accessible through a context-optimized layer that agents can better understand and operate end-to-end. GitHub: Key Benchmarks (vs. Supabase MCP): - 14% higher accuracy - 1.3x faster per task - 2.4x fewer tokens Better. Faster. Cheaper. Build features more quickly and confidently — all at 41.7% of the cost. Shipping your ideas today. $ npx @insforge/cli create

Introducing InsForge 2.0: The Backend for Agentic Development Our OSS backend provides databases, auth, storage, model gateway, and edge functions accessible through a context-optimized layer that agents can better understand and operate end-to-end. GitHub: Key Benchmarks (vs. Supabase MCP): - 14% higher accuracy - 1.3x faster per task - 2.4x fewer tokens Better. Faster. Cheaper. Build features more quickly and confidently — all at 41.7% of the cost. Shipping your ideas today. $ npx @insforge/cli create

InsForge

1,592,962 views • 7 months ago

Introducing InsForge Payments. Your coding agent can now set up Stripe payments for your app using InsForge. Describe your pricing and watch it create products and prices in Stripe, add checkout, and set up billing portal sessions. Test first. Go live when ready.

Introducing InsForge Payments. Your coding agent can now set up Stripe payments for your app using InsForge. Describe your pricing and watch it create products and prices in Stripe, add checkout, and set up billing portal sessions. Test first. Go live when ready.

InsForge

674,924 views • 4 months ago

i just built a 4-agent software team. everything runs from Telegram and gets managed on a kanban board. a project manager who plans the work, a backend developer, a frontend developer, and a tester. the PM reads a goal, breaks it into linked tasks, and assigns each to the right agent. the thing that makes them a team instead of four strangers is a shared kanban board. every task is a row that survives crashes, and when an agent finishes, it writes a summary of what it built and what the next agent needs to know. the next agent reads that summary before it starts. so the frontend developer never has to guess the API shape, and the tester knows exactly what to verify. the hardest part was not the coordination. it was building an agent that could actually act like a backend engineer. a backend engineer stands up a database, wires auth, manages storage, deploys functions, and keeps all of it consistent while the rest of the team builds on top. an agent doing this from scratch drowns. it burns its context window remembering which tables exist and which endpoint it created three steps ago, and the work degrades fast. so the backend agent needs a backend built for agents, not for humans clicking through a dashboard. that is where InsForge came in. it is an open-source, agent-native backend, and i added it to my backend developer agent as a skill. a skill is a step-by-step guide that teaches the agent how to do a specific kind of work. with InsForge installed, the agent stopped improvising infrastructure and followed a reliable path: create the project, define the database, set up auth, deploy functions. to test the whole team, i had them build a working Google Docs clone, AI features included. the backend agent spun up the full service on its own. database tables, user auth, document handling, and edge functions running real TypeScript, all in one dashboard. the frontend agent read that summary and built the UI on top of it, and the tester closed the loop. the result was a backend an agent could reason about end to end, instead of one it kept getting lost inside. if you are building an AI backend engineer, InsForge is worth a look, it's 100% open-source. InsForge GitHub: (don't forget to star 🌟) the full article on Hermes Kanban: Mission Control for your Agents is quoted below.

i just built a 4-agent software team. everything runs from Telegram and gets managed on a kanban board. a project manager who plans the work, a backend developer, a frontend developer, and a tester. the PM reads a goal, breaks it into linked tasks, and assigns each to the right agent. the thing that makes them a team instead of four strangers is a shared kanban board. every task is a row that survives crashes, and when an agent finishes, it writes a summary of what it built and what the next agent needs to know. the next agent reads that summary before it starts. so the frontend developer never has to guess the API shape, and the tester knows exactly what to verify. the hardest part was not the coordination. it was building an agent that could actually act like a backend engineer. a backend engineer stands up a database, wires auth, manages storage, deploys functions, and keeps all of it consistent while the rest of the team builds on top. an agent doing this from scratch drowns. it burns its context window remembering which tables exist and which endpoint it created three steps ago, and the work degrades fast. so the backend agent needs a backend built for agents, not for humans clicking through a dashboard. that is where InsForge came in. it is an open-source, agent-native backend, and i added it to my backend developer agent as a skill. a skill is a step-by-step guide that teaches the agent how to do a specific kind of work. with InsForge installed, the agent stopped improvising infrastructure and followed a reliable path: create the project, define the database, set up auth, deploy functions. to test the whole team, i had them build a working Google Docs clone, AI features included. the backend agent spun up the full service on its own. database tables, user auth, document handling, and edge functions running real TypeScript, all in one dashboard. the frontend agent read that summary and built the UI on top of it, and the tester closed the loop. the result was a backend an agent could reason about end to end, instead of one it kept getting lost inside. if you are building an AI backend engineer, InsForge is worth a look, it's 100% open-source. InsForge GitHub: (don't forget to star 🌟) the full article on Hermes Kanban: Mission Control for your Agents is quoted below.

Akshay 🚀

123,101 views • 4 months ago

Finally, I'm done with the basics of backend engineering on the all-in-one resources to learn backend engineering I'm working on This basic section includes - Internet - HTTP - Servers - Web Dev fundamentals - Fundamentals of Operating Systems - A server-side language -> JavaScript | Go | Rust | Node - A framework -> Express | Nestjs | Django | Laravel | Spring Boot This is the boring part. Don't worry. We are getting into the fun part soon. Still working on it, updating it, Adding more, and making it better for you. Follow me. Retweet. Comment "Backend." I will DM you the PDF version when it's out. Enjoy the web version, and send in your feedback on how we can improve it for everyone.

Finally, I'm done with the basics of backend engineering on the all-in-one resources to learn backend engineering I'm working on This basic section includes - Internet - HTTP - Servers - Web Dev fundamentals - Fundamentals of Operating Systems - A server-side language -> JavaScript | Go | Rust | Node - A framework -> Express | Nestjs | Django | Laravel | Spring Boot This is the boring part. Don't worry. We are getting into the fun part soon. Still working on it, updating it, Adding more, and making it better for you. Follow me. Retweet. Comment "Backend." I will DM you the PDF version when it's out. Enjoy the web version, and send in your feedback on how we can improve it for everyone.

Solomon Eseme

92,216 views • 3 years ago

We’re building the agent-native alternative to AWS, but it's actually good. - 33,000+ projects created (10% WoW growth) - 2x weekly active projects (2000 → 4000+) - 11,000+ GitHub stars (5x in the last 3 months) - faster than React, Supabase, and Linux Introducing InsForge ↓

We’re building the agent-native alternative to AWS, but it's actually good. - 33,000+ projects created (10% WoW growth) - 2x weekly active projects (2000 → 4000+) - 11,000+ GitHub stars (5x in the last 3 months) - faster than React, Supabase, and Linux Introducing InsForge ↓

InsForge

603,836 views • 4 months ago

We just launched one-click export to Lovable! ❤️ Lovable is an incredible AI app builder that turns your ideas into real, working software—including the frontend, backend, and database. When you export your Stitch project: ✨ Lovable opens automatically ✨ Stitch passes over your screenshots ✨ We include exact references to the HTML right in the prompt Lovable downloads your HTML and creates a pixel-perfect, full-stack application based on your exact aesthetic. Have a blast turning your designs into reality! 🚀

We just launched one-click export to Lovable! ❤️ Lovable is an incredible AI app builder that turns your ideas into real, working software—including the frontend, backend, and database. When you export your Stitch project: ✨ Lovable opens automatically ✨ Stitch passes over your screenshots ✨ We include exact references to the HTML right in the prompt Lovable downloads your HTML and creates a pixel-perfect, full-stack application based on your exact aesthetic. Have a blast turning your designs into reality! 🚀

Stitch by Google

64,009 views • 5 months ago

Today we’re launching 4 open-source Next.js app templates: 1. Basic starter 2. Chatbot 3. CRM 4. Ecommerce We made building real full-stack apps even easier on InsForge. Try one in your next project: npx @insforge/cli create Thread below with all 4 repos ↓

Today we’re launching 4 open-source Next.js app templates: 1. Basic starter 2. Chatbot 3. CRM 4. Ecommerce We made building real full-stack apps even easier on InsForge. Try one in your next project: npx @insforge/cli create Thread below with all 4 repos ↓

InsForge

671,992 views • 6 months ago

Imagine a plane crashing and only 3% of people on the aircraft know where the emergency exits are. We need to work towards taking these backend features right into the frontend so all users can access them. An exit is only fair when it is visible to all. (Re: L2 exit to L1)

Imagine a plane crashing and only 3% of people on the aircraft know where the emergency exits are. We need to work towards taking these backend features right into the frontend so all users can access them. An exit is only fair when it is visible to all. (Re: L2 exit to L1)

binji

13,920 views • 3 months ago

Claude Code tip: stop burning Opus 5.5 context on tasks Sonnet 5.5 can swarm, while Fable 5.1 sits idle spin up an autonomous team with --agent claude --agent architect Opus 5.5 acts as the lead architect, scoping the work and validating every worktree before it merges Sonnet 5.5 teammates take isolated git worktrees and hammer out implementations at medium effort Fable 5.1 sits on the team as a dedicated adversary, only stepping in at three critical moments: → before a contract locks: does the frontend payload match the backend schema? → when a test breaks twice: are we patching the bug or just hiding the symptom? → before the merge: what attack surface or edge case did everyone overlook? Sonnet 5.5 builds. Fable 5.1 challenges. Opus 5.5 merges and ships Jev engineering runs the identical pattern one layer down: mechanical decisions that need no reasoning (which file to open, which tool to invoke, retry or abort) execute in 16ms, so the frontier models only wake up when execution paths actually diverge Plan on high. Delegate on medium. Keep the adversary on peer-to-peer call. - the full team setup > architect runs the main session on Opus 5.5 at high effort > teammate-ux builds client state inside its own git worktree > teammate-back writes endpoints and migrations inside a separate worktree > teammate-adv runs on Fable 5.1, read-only, and attacks every diff > teammates message each other by name without routing through the lead > tools: Agent(ux, backend, adversary) locks the architect to spawning only its own team paste the setup and this prompt into Claude Code below: "Configure my Claude Code workspace for an autonomous agent team: 1. Audit ~/.claude/agents and .claude/agents for definitions fitting architect, ux, backend and adversary > Generate definitions only for missing roles > Pin model: sonnet, effort: medium, isolation: worktree for ux and backend, and model: fable with read-only tools for adversary > Give architect model: opus, effort: high and tools: Agent(ux, backend, adversary) 2. Check ~/.claude/settings.json and my env: > Report CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS if it's set, since it turns named subagents into teammates without worktrees > Report CLAUDE_CODE_SUBAGENT_MODEL_FORCE and CLAUDE_CODE_EFFORT_LEVEL, which can override the pinned models and effort 3. Write each role's prompt so ux and backend own separate folders, message each other by name when a contract changes, and hand every diff to adversary before reporting back 4. Add a team rule to ~/.claude/CLAUDE.md: no worktree merges until adversary has reviewed it Show every configuration diff first. Do not apply edits until confirmed" save it

Claude Code tip: stop burning Opus 5.5 context on tasks Sonnet 5.5 can swarm, while Fable 5.1 sits idle spin up an autonomous team with --agent claude --agent architect Opus 5.5 acts as the lead architect, scoping the work and validating every worktree before it merges Sonnet 5.5 teammates take isolated git worktrees and hammer out implementations at medium effort Fable 5.1 sits on the team as a dedicated adversary, only stepping in at three critical moments: → before a contract locks: does the frontend payload match the backend schema? → when a test breaks twice: are we patching the bug or just hiding the symptom? → before the merge: what attack surface or edge case did everyone overlook? Sonnet 5.5 builds. Fable 5.1 challenges. Opus 5.5 merges and ships Jev engineering runs the identical pattern one layer down: mechanical decisions that need no reasoning (which file to open, which tool to invoke, retry or abort) execute in 16ms, so the frontier models only wake up when execution paths actually diverge Plan on high. Delegate on medium. Keep the adversary on peer-to-peer call. - the full team setup > architect runs the main session on Opus 5.5 at high effort > teammate-ux builds client state inside its own git worktree > teammate-back writes endpoints and migrations inside a separate worktree > teammate-adv runs on Fable 5.1, read-only, and attacks every diff > teammates message each other by name without routing through the lead > tools: Agent(ux, backend, adversary) locks the architect to spawning only its own team paste the setup and this prompt into Claude Code below: "Configure my Claude Code workspace for an autonomous agent team: 1. Audit ~/.claude/agents and .claude/agents for definitions fitting architect, ux, backend and adversary > Generate definitions only for missing roles > Pin model: sonnet, effort: medium, isolation: worktree for ux and backend, and model: fable with read-only tools for adversary > Give architect model: opus, effort: high and tools: Agent(ux, backend, adversary) 2. Check ~/.claude/settings.json and my env: > Report CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS if it's set, since it turns named subagents into teammates without worktrees > Report CLAUDE_CODE_SUBAGENT_MODEL_FORCE and CLAUDE_CODE_EFFORT_LEVEL, which can override the pinned models and effort 3. Write each role's prompt so ux and backend own separate folders, message each other by name when a contract changes, and hand every diff to adversary before reporting back 4. Add a team rule to ~/.claude/CLAUDE.md: no worktree merges until adversary has reviewed it Show every configuration diff first. Do not apply edits until confirmed" save it

Khairallah AL-Awady

36,814 views • 16 days ago

HIS VIBE CODING APP WAS HACKED and he almost lost $10,000 because of it After analyzing why this happened, we managed to find 3 most common mistakes that appear in the majority of vibe coding projects: • Misconfigured RLS • No backend rate limits • AI calls made from the frontend And vibe coding is actually not dangerous, you just need to pay more attention to these points when creating a project: • Protecting access to data • Limiting resource usage • Hiding all secrets on the backend • Controlling costs • Regularly doing security reviews It’s also important from the very beginning to give AI tools for improving code security And I talked about them in the article below👇

HIS VIBE CODING APP WAS HACKED and he almost lost $10,000 because of it After analyzing why this happened, we managed to find 3 most common mistakes that appear in the majority of vibe coding projects: • Misconfigured RLS • No backend rate limits • AI calls made from the frontend And vibe coding is actually not dangerous, you just need to pay more attention to these points when creating a project: • Protecting access to data • Limiting resource usage • Hiding all secrets on the backend • Controlling costs • Regularly doing security reviews It’s also important from the very beginning to give AI tools for improving code security And I talked about them in the article below👇

TolyaDV

45,570 views • 4 months ago

Long, long, ago, coding agents could do nothing but write code Forger was powerless, just like all the others. Claude Code. Codex. Cursor. Until one day... InsForge Launch Week 2 starts Monday, June 1st

Long, long, ago, coding agents could do nothing but write code Forger was powerless, just like all the others. Claude Code. Codex. Cursor. Until one day... InsForge Launch Week 2 starts Monday, June 1st

InsForge

35,357 views • 4 months ago

$I don't understand why everyone isn't doing this yet. Anthropic released a guide for Claude that makes work much more efficient while burning a fraction of the tokens Main Idea: Agent Teams in Claude Code Instead of burning the expensive Opus 5.5 context on every line of code, tasks are delegated across an autonomous swarm: Architect (Opus 5.5, high effort): Plans the structure and handles final PR merges Developers (Sonnet 5.5, medium effort): Write code inside isolated ⁠git worktree⁠ environments (UI/UX, backend) Adversary / Critic (Fable 5.1): Steps in at contract boundaries, recurring test failures, and pre-PR audits to catch vulnerabilities Script Layer (Jev): Handles mechanical tasks (opening files, CLI invocations) in 16ms without involving LLMs Launch the team via terminal: claude --teammates ux,backend,adversary Configure my Claude Code workspace for autonomous agent teams: In ⁠~/.claude/teams⁠, generate missing roles: set workers to ⁠Sonnet⁠ (⁠medium effort⁠) in separate ⁠git worktree⁠ branches, and the adversary to ⁠Fable⁠ In ⁠~/.claude/settings.json⁠, set ⁠CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1⁠ and lock the main session to ⁠effortLevel: high⁠ Initialize ⁠tasks.json⁠ with file-level locking for direct peer coordination In ⁠CLAUDE.md⁠, add a rule requiring an adversary review before opening a PR. Show configuration diffs before applying edits >$

I don't understand why everyone isn't doing this yet. Anthropic released a guide for Claude that makes work much more efficient while burning a fraction of the tokens Main Idea: Agent Teams in Claude Code Instead of burning the expensive Opus 5.5 context on every line of code, tasks are delegated across an autonomous swarm: Architect (Opus 5.5, high effort): Plans the structure and handles final PR merges Developers (Sonnet 5.5, medium effort): Write code inside isolated ⁠git worktree⁠ environments (UI/UX, backend) Adversary / Critic (Fable 5.1): Steps in at contract boundaries, recurring test failures, and pre-PR audits to catch vulnerabilities Script Layer (Jev): Handles mechanical tasks (opening files, CLI invocations) in 16ms without involving LLMs Launch the team via terminal: claude --teammates ux,backend,adversary Configure my Claude Code workspace for autonomous agent teams: In ⁠~/.claude/teams⁠, generate missing roles: set workers to ⁠Sonnet⁠ (⁠medium effort⁠) in separate ⁠git worktree⁠ branches, and the adversary to ⁠Fable⁠ In ⁠~/.claude/settings.json⁠, set ⁠CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1⁠ and lock the main session to ⁠effortLevel: high⁠ Initialize ⁠tasks.json⁠ with file-level locking for direct peer coordination In ⁠CLAUDE.md⁠, add a rule requiring an adversary review before opening a PR. Show configuration diffs before applying edits >

Bober_smart

333,966 views • 19 days ago

this GPT-6 Astra setup is getting f...cking dangerous 8 GitHub repos for turning one coding agent into an entire engineering team 01 OpenHands ▸ → autonomous software development 02 SWE-agent ▸ → issue → inspect repo → patch → test BUILD THE WORKERS 03 Aider ▸ → pair programming directly inside your repo 04 Cline ▸ → plan → edit → terminal → browser 05 Roo Code ▸ → specialized coding agent modes GIVE THEM A WORKSPACE 06 E2B ▸ → isolated sandboxes for parallel execution 07 Daytona ▸ → reproducible environments for agent workloads CHECK THE WORK 08 SWE-bench ▸ → test agents against real GitHub issues the swarm: issue → planner → split tasks → parallel agents → patches → tests → critic → merge one agent fixes the backend another handles the frontend another runs tests while a critic checks the diff that's where GPT-6 Astra gets interesting not one AI writing more code an entire software team running inside the repo ⭣

this GPT-6 Astra setup is getting f...cking dangerous 8 GitHub repos for turning one coding agent into an entire engineering team 01 OpenHands ▸ → autonomous software development 02 SWE-agent ▸ → issue → inspect repo → patch → test BUILD THE WORKERS 03 Aider ▸ → pair programming directly inside your repo 04 Cline ▸ → plan → edit → terminal → browser 05 Roo Code ▸ → specialized coding agent modes GIVE THEM A WORKSPACE 06 E2B ▸ → isolated sandboxes for parallel execution 07 Daytona ▸ → reproducible environments for agent workloads CHECK THE WORK 08 SWE-bench ▸ → test agents against real GitHub issues the swarm: issue → planner → split tasks → parallel agents → patches → tests → critic → merge one agent fixes the backend another handles the frontend another runs tests while a critic checks the diff that's where GPT-6 Astra gets interesting not one AI writing more code an entire software team running inside the repo ⭣

Lummox

35,870 views • 21 days ago

QVAC SDK 0.15.0 is live. This release adds multiple prompts batching, brings a native AMD GPU backend to the stack, moves more vision encoders onto mobile GPUs, and adds a second local coding-agent integration. Main highlights: - Prompt batching for the LLM addon. Batch multiple prompts into one job and process them concurrently, with each answer returned the moment its generation finishes. - Native AMD GPU backend. A first-class HIP/ROCm backend in @qvac/vla-ggml, auto-selected over Vulkan with clean fallback when ROCm is absent. - A second local coding agent. OpenClaw joins OpenCode for local, cloud-free agent workflows. AGENTS - OpenCode plugin update (@qvac/opencode-plugin). Aligned with the current SDK, CLI, and AI SDK provider packages. A fresh install runs OpenCode against managed local QVAC models out of the box, from the default qvac/qwen3.5-9b, with no manual qvac serve setup. - OpenClaw plugin (@qvac/openclaw-plugin). A second coding-agent integration alongside OpenCode. A fresh setup installs the plugin, creates a local qvac provider through onboarding, and runs a QVAC model through OpenClaw🦞's local service path. LANGUAGE MODELS - Prompt batching (LLM addon). Batch multiple prompts in one job and run them concurrently, each answer returns the moment its generation finishes, no waiting on the others. - Reasoning-context trimming on hybrid + recurrent models (@qvac/llm-llamacpp). remove_thinking_from_context now works beyond pure-attention models. Same JS API, no throw. VOICE AND SPEECH - Transcription (transcription-parakeet 0.9.0). More robust CPU fallback on GPU failure and a faster Vulkan backend on Pixel 9. - Text-to-speech features (tts-ggml 0.4.0). Adds LavaSR for noise removal and adjustable output frequency up to 48 kHz, plus Japanese via Chatterbox. - Text-to-speech fixes (tts-ggml 0.4.1). CPU fallback on GPU failure, a q8_0 KV crash fix on Metal with Chatterbox. VISION - Qwen3.5 vision encoder on GPU (Android). Image encoder moves onto the phone GPU, with a smarter tile-grid preprocessor and default image-token caps, for flagship Android: Vulkan on Mali (Pixel 9 Pro) and OpenCL on Adreno 830 (Galaxy S25). - Gemma-4 vision encoder on GPU (Android). Vision encoder runs on the phone GPU instead of CPU, same flagship Android targets. PLATFORM AND PERFORMANCE - AMD GPU backend (@qvac/vla-ggml). Native HIP/ROCm backend, auto-selected over Vulkan with clean fallback when ROCm is absent (Linux x64 only). Comes with ~23% faster than Vulkan, ~14% faster than PyTorch-ROCm, parity preserved. Unified code style. A cleaner, more consistent, easier-to-contribute codebase. Let's build. npm install @qvac/sdk

QVAC SDK 0.15.0 is live. This release adds multiple prompts batching, brings a native AMD GPU backend to the stack, moves more vision encoders onto mobile GPUs, and adds a second local coding-agent integration. Main highlights: - Prompt batching for the LLM addon. Batch multiple prompts into one job and process them concurrently, with each answer returned the moment its generation finishes. - Native AMD GPU backend. A first-class HIP/ROCm backend in @qvac/vla-ggml, auto-selected over Vulkan with clean fallback when ROCm is absent. - A second local coding agent. OpenClaw joins OpenCode for local, cloud-free agent workflows. AGENTS - OpenCode plugin update (@qvac/opencode-plugin). Aligned with the current SDK, CLI, and AI SDK provider packages. A fresh install runs OpenCode against managed local QVAC models out of the box, from the default qvac/qwen3.5-9b, with no manual qvac serve setup. - OpenClaw plugin (@qvac/openclaw-plugin). A second coding-agent integration alongside OpenCode. A fresh setup installs the plugin, creates a local qvac provider through onboarding, and runs a QVAC model through OpenClaw🦞's local service path. LANGUAGE MODELS - Prompt batching (LLM addon). Batch multiple prompts in one job and run them concurrently, each answer returns the moment its generation finishes, no waiting on the others. - Reasoning-context trimming on hybrid + recurrent models (@qvac/llm-llamacpp). remove_thinking_from_context now works beyond pure-attention models. Same JS API, no throw. VOICE AND SPEECH - Transcription (transcription-parakeet 0.9.0). More robust CPU fallback on GPU failure and a faster Vulkan backend on Pixel 9. - Text-to-speech features (tts-ggml 0.4.0). Adds LavaSR for noise removal and adjustable output frequency up to 48 kHz, plus Japanese via Chatterbox. - Text-to-speech fixes (tts-ggml 0.4.1). CPU fallback on GPU failure, a q8_0 KV crash fix on Metal with Chatterbox. VISION - Qwen3.5 vision encoder on GPU (Android). Image encoder moves onto the phone GPU, with a smarter tile-grid preprocessor and default image-token caps, for flagship Android: Vulkan on Mali (Pixel 9 Pro) and OpenCL on Adreno 830 (Galaxy S25). - Gemma-4 vision encoder on GPU (Android). Vision encoder runs on the phone GPU instead of CPU, same flagship Android targets. PLATFORM AND PERFORMANCE - AMD GPU backend (@qvac/vla-ggml). Native HIP/ROCm backend, auto-selected over Vulkan with clean fallback when ROCm is absent (Linux x64 only). Comes with ~23% faster than Vulkan, ~14% faster than PyTorch-ROCm, parity preserved. Unified code style. A cleaner, more consistent, easier-to-contribute codebase. Let's build. npm install @qvac/sdk

QVAC

29,260,217 views • 3 months ago

Introducing... MERGE MOMMY my eve PR review + auto approve bot Like all of you, I was drowning in pull requests, so I built a little agent to risk score PRs and auto approve low risk ones, inspired by previous How I AI guest Brian Scanlan and the work of the Intercom team In this mini ep, I teach you 02:35 why you shouldn't approve every PR by hand 06:10 how the eve agent framework works 11:36 designing your PR review agent 13:07 setting up + deploying eve w codex browser use 17:23 Merge Mommy in action I recommend this to every eng team out there -- if you haven't built one yet, today is your day. As always, huge thanks to our amazing sponsor WorkOS - Make your app Enterprise Ready today

Introducing... MERGE MOMMY my eve PR review + auto approve bot Like all of you, I was drowning in pull requests, so I built a little agent to risk score PRs and auto approve low risk ones, inspired by previous How I AI guest Brian Scanlan and the work of the Intercom team In this mini ep, I teach you 02:35 why you shouldn't approve every PR by hand 06:10 how the eve agent framework works 11:36 designing your PR review agent 13:07 setting up + deploying eve w codex browser use 17:23 Merge Mommy in action I recommend this to every eng team out there -- if you haven't built one yet, today is your day. As always, huge thanks to our amazing sponsor WorkOS - Make your app Enterprise Ready today

claire vo 🖤

25,632 views • 2 months ago

10 GitHub repos with ⭐ 1.2M+ combined stars that help you build your AI agent from scratch 1. n8n 201k⭐ - visual agent builder with 400+ integrations. Wire up multi-step AI workflows without writing a backend. 2. Hermes Agent 231k⭐ - the fastest-growing open-source agent framework of 2026. Self-improving skill loop, persistent memory across sessions. 3. Dify 152k⭐ - visual builder for agentic workflows and RAG pipelines, production-ready out of the box. 4. LangChain 144k⭐ - the standard framework for LLM apps. Chains, agents, retrieval, all in one place. 5. MetaGPT 70k⭐ - simulates an entire software company. PM, architect, engineer agents that build your app together. 6. AutoGen 60k⭐ - Microsoft's framework for multi-agent conversations and coordination. 7. OpenCode 198k⭐ - terminal coding agent, provider-agnostic across 75+ models. The go-to harness for agentic coding in 2026. 8. LlamaIndex 52k⭐ - connects your agents to your actual data. Indexing and retrieval done right. 9. CrewAI 57k⭐ - role-based crews of agents. Reads almost like plain English, no graph to learn. 10. LangGraph 40k⭐ - graph-based orchestration when you need precise control over agent state in production. Save this before you pay for another agent-building course 👇

10 GitHub repos with ⭐ 1.2M+ combined stars that help you build your AI agent from scratch 1. n8n 201k⭐ - visual agent builder with 400+ integrations. Wire up multi-step AI workflows without writing a backend. 2. Hermes Agent 231k⭐ - the fastest-growing open-source agent framework of 2026. Self-improving skill loop, persistent memory across sessions. 3. Dify 152k⭐ - visual builder for agentic workflows and RAG pipelines, production-ready out of the box. 4. LangChain 144k⭐ - the standard framework for LLM apps. Chains, agents, retrieval, all in one place. 5. MetaGPT 70k⭐ - simulates an entire software company. PM, architect, engineer agents that build your app together. 6. AutoGen 60k⭐ - Microsoft's framework for multi-agent conversations and coordination. 7. OpenCode 198k⭐ - terminal coding agent, provider-agnostic across 75+ models. The go-to harness for agentic coding in 2026. 8. LlamaIndex 52k⭐ - connects your agents to your actual data. Indexing and retrieval done right. 9. CrewAI 57k⭐ - role-based crews of agents. Reads almost like plain English, no graph to learn. 10. LangGraph 40k⭐ - graph-based orchestration when you need precise control over agent state in production. Save this before you pay for another agent-building course 👇

unicode

30,511 views • 2 months ago