Epic Vibecoder
@epicvibecoder.bsky.social
AI tool drops, daily 🛠️ underground gems + the news that matters. no paywalls. → t.me/epicvibecoder
one video in, dubbed and subtitled video out.
krillinai: download, transcribe, translate, voice clone dub and vertical recut in one pipeline. go, gpl-3.0, self hosted.
which language would double your audience?
github.com/krillinai/Kr...
krillinai: download, transcribe, translate, voice clone dub and vertical recut in one pipeline. go, gpl-3.0, self hosted.
which language would double your audience?
github.com/krillinai/Kr...
August 18, 2026 at 7:01 PM
one video in, dubbed and subtitled video out.
krillinai: download, transcribe, translate, voice clone dub and vertical recut in one pipeline. go, gpl-3.0, self hosted.
which language would double your audience?
github.com/krillinai/Kr...
krillinai: download, transcribe, translate, voice clone dub and vertical recut in one pipeline. go, gpl-3.0, self hosted.
which language would double your audience?
github.com/krillinai/Kr...
the guard dog that stops your agent dropping the database.
doberman gates every agent input, output and tool call at runtime. unsafe actions get blocked before they run, not after. apache-2.0.
what would you never let an agent run?
github.com/fu351/Doberm...
doberman gates every agent input, output and tool call at runtime. unsafe actions get blocked before they run, not after. apache-2.0.
what would you never let an agent run?
github.com/fu351/Doberm...
August 18, 2026 at 5:30 PM
the guard dog that stops your agent dropping the database.
doberman gates every agent input, output and tool call at runtime. unsafe actions get blocked before they run, not after. apache-2.0.
what would you never let an agent run?
github.com/fu351/Doberm...
doberman gates every agent input, output and tool call at runtime. unsafe actions get blocked before they run, not after. apache-2.0.
what would you never let an agent run?
github.com/fu351/Doberm...
drag the box and the mermaid code rewrites itself.
visimer is a visual editor for mermaid that keeps the source in sync, so your diagram stays a file you can diff. mit, shipped this month.
how many diagrams did you abandon mid-edit?
github.com/inkeep/visimer
visimer is a visual editor for mermaid that keeps the source in sync, so your diagram stays a file you can diff. mit, shipped this month.
how many diagrams did you abandon mid-edit?
github.com/inkeep/visimer
August 18, 2026 at 3:30 PM
drag the box and the mermaid code rewrites itself.
visimer is a visual editor for mermaid that keeps the source in sync, so your diagram stays a file you can diff. mit, shipped this month.
how many diagrams did you abandon mid-edit?
github.com/inkeep/visimer
visimer is a visual editor for mermaid that keeps the source in sync, so your diagram stays a file you can diff. mit, shipped this month.
how many diagrams did you abandon mid-edit?
github.com/inkeep/visimer
switch coding agents without losing the plot.
ai-memory is a long term memory layer that lives outside your cli agents, so context survives the vendor swap. rust, mit.
which agent would you drop if moving out was free?
github.com/akitaonrails...
ai-memory is a long term memory layer that lives outside your cli agents, so context survives the vendor swap. rust, mit.
which agent would you drop if moving out was free?
github.com/akitaonrails...
August 18, 2026 at 12:00 PM
switch coding agents without losing the plot.
ai-memory is a long term memory layer that lives outside your cli agents, so context survives the vendor swap. rust, mit.
which agent would you drop if moving out was free?
github.com/akitaonrails...
ai-memory is a long term memory layer that lives outside your cli agents, so context survives the vendor swap. rust, mit.
which agent would you drop if moving out was free?
github.com/akitaonrails...
it reviews the pull request before your teammate has to.
pr-af reads the whole diff and leaves comments worth acting on. tops the open code review benchmark, runs in your own ci.
why did your last pr sit for three days?
github.com/Agent-Field/...
pr-af reads the whole diff and leaves comments worth acting on. tops the open code review benchmark, runs in your own ci.
why did your last pr sit for three days?
github.com/Agent-Field/...
August 18, 2026 at 8:30 AM
it reviews the pull request before your teammate has to.
pr-af reads the whole diff and leaves comments worth acting on. tops the open code review benchmark, runs in your own ci.
why did your last pr sit for three days?
github.com/Agent-Field/...
pr-af reads the whole diff and leaves comments worth acting on. tops the open code review benchmark, runs in your own ci.
why did your last pr sit for three days?
github.com/Agent-Field/...
a real cad app that opens dwg and boots instantly.
opencadstudio: 2d and 3d cad in rust, gpu accelerated, reads dwg and dxf. gpl-3.0, no subscription.
which paid tool would you drop if file formats let you?
github.com/HakanSeven12...
opencadstudio: 2d and 3d cad in rust, gpu accelerated, reads dwg and dxf. gpl-3.0, no subscription.
which paid tool would you drop if file formats let you?
github.com/HakanSeven12...
August 17, 2026 at 8:30 AM
a real cad app that opens dwg and boots instantly.
opencadstudio: 2d and 3d cad in rust, gpu accelerated, reads dwg and dxf. gpl-3.0, no subscription.
which paid tool would you drop if file formats let you?
github.com/HakanSeven12...
opencadstudio: 2d and 3d cad in rust, gpu accelerated, reads dwg and dxf. gpl-3.0, no subscription.
which paid tool would you drop if file formats let you?
github.com/HakanSeven12...
your chat history is a straight line. your thinking is not.
thoughtdag turns llm conversations into an editable graph on an infinite canvas. branch, prune, rewire.
how much thinking have you lost to scrollback?
github.com/chenxiachan/...
thoughtdag turns llm conversations into an editable graph on an infinite canvas. branch, prune, rewire.
how much thinking have you lost to scrollback?
github.com/chenxiachan/...
August 16, 2026 at 7:00 PM
your chat history is a straight line. your thinking is not.
thoughtdag turns llm conversations into an editable graph on an infinite canvas. branch, prune, rewire.
how much thinking have you lost to scrollback?
github.com/chenxiachan/...
thoughtdag turns llm conversations into an editable graph on an infinite canvas. branch, prune, rewire.
how much thinking have you lost to scrollback?
github.com/chenxiachan/...
it reads the manga panel, wipes the text, redraws it in english.
koharu: bubble detection, ocr, inpainting and typesetting in one rust pipeline. runs fully local.
what would you translate first?
github.com/mayocream/ko...
koharu: bubble detection, ocr, inpainting and typesetting in one rust pipeline. runs fully local.
what would you translate first?
github.com/mayocream/ko...
August 16, 2026 at 5:34 PM
it reads the manga panel, wipes the text, redraws it in english.
koharu: bubble detection, ocr, inpainting and typesetting in one rust pipeline. runs fully local.
what would you translate first?
github.com/mayocream/ko...
koharu: bubble detection, ocr, inpainting and typesetting in one rust pipeline. runs fully local.
what would you translate first?
github.com/mayocream/ko...
git, but for the agents editing your code.
atlas tracks every edit per agent and lets you query all of them in one place. mit, three months old.
do you actually know what your agents changed?
github.com/pacifio/atlas
atlas tracks every edit per agent and lets you query all of them in one place. mit, three months old.
do you actually know what your agents changed?
github.com/pacifio/atlas
August 16, 2026 at 3:30 PM
git, but for the agents editing your code.
atlas tracks every edit per agent and lets you query all of them in one place. mit, three months old.
do you actually know what your agents changed?
github.com/pacifio/atlas
atlas tracks every edit per agent and lets you query all of them in one place. mit, three months old.
do you actually know what your agents changed?
github.com/pacifio/atlas
point a coding agent at your repo and let it hunt bugs.
deepsec is a security harness from vercel labs. it follows the attacker's path, not a pattern list.
would you let an agent audit your prod repo?
github.com/vercel-labs/...
deepsec is a security harness from vercel labs. it follows the attacker's path, not a pattern list.
would you let an agent audit your prod repo?
github.com/vercel-labs/...
August 16, 2026 at 12:00 PM
point a coding agent at your repo and let it hunt bugs.
deepsec is a security harness from vercel labs. it follows the attacker's path, not a pattern list.
would you let an agent audit your prod repo?
github.com/vercel-labs/...
deepsec is a security harness from vercel labs. it follows the attacker's path, not a pattern list.
would you let an agent audit your prod repo?
github.com/vercel-labs/...
write lua, get a printable 3d part.
luacad is parametric cad scripted in lua on top of openscad. export stl, dxf, svg, obj. your geometry becomes a file you can diff and loop over.
if your cad lived in git, what would you design differently?
github.com/ad-si/LuaCAD
luacad is parametric cad scripted in lua on top of openscad. export stl, dxf, svg, obj. your geometry becomes a file you can diff and loop over.
if your cad lived in git, what would you design differently?
github.com/ad-si/LuaCAD
August 15, 2026 at 7:00 PM
write lua, get a printable 3d part.
luacad is parametric cad scripted in lua on top of openscad. export stl, dxf, svg, obj. your geometry becomes a file you can diff and loop over.
if your cad lived in git, what would you design differently?
github.com/ad-si/LuaCAD
luacad is parametric cad scripted in lua on top of openscad. export stl, dxf, svg, obj. your geometry becomes a file you can diff and loop over.
if your cad lived in git, what would you design differently?
github.com/ad-si/LuaCAD
a research agent that cannot lie about its sources.
mole is a terminal deep-research agent in go: enforced token budget, verified quotes, and a privacy boundary for local data.
how much per run before you want a hard cap?
github.com/lajosdeme/mole
mole is a terminal deep-research agent in go: enforced token budget, verified quotes, and a privacy boundary for local data.
how much per run before you want a hard cap?
github.com/lajosdeme/mole
August 15, 2026 at 5:30 PM
a research agent that cannot lie about its sources.
mole is a terminal deep-research agent in go: enforced token budget, verified quotes, and a privacy boundary for local data.
how much per run before you want a hard cap?
github.com/lajosdeme/mole
mole is a terminal deep-research agent in go: enforced token budget, verified quotes, and a privacy boundary for local data.
how much per run before you want a hard cap?
github.com/lajosdeme/mole
your team's agents finally share one brain.
a team-level memory hub that turns chats, docs and code into four reusable assets every agent can read, across frameworks.
is your agent memory shared, or private per dev?
github.com/TencentCloud...
a team-level memory hub that turns chats, docs and code into four reusable assets every agent can read, across frameworks.
is your agent memory shared, or private per dev?
github.com/TencentCloud...
August 15, 2026 at 3:30 PM
your team's agents finally share one brain.
a team-level memory hub that turns chats, docs and code into four reusable assets every agent can read, across frameworks.
is your agent memory shared, or private per dev?
github.com/TencentCloud...
a team-level memory hub that turns chats, docs and code into four reusable assets every agent can read, across frameworks.
is your agent memory shared, or private per dev?
github.com/TencentCloud...
an agent that rewrites its own playbook between runs.
prime agent is a self-improving rlm agent for long autonomous coding work. it revises its own policy after each run. no cold starts.
trust an agent that edits itself?
github.com/PrimeIntelle...
prime agent is a self-improving rlm agent for long autonomous coding work. it revises its own policy after each run. no cold starts.
trust an agent that edits itself?
github.com/PrimeIntelle...
August 15, 2026 at 1:00 PM
an agent that rewrites its own playbook between runs.
prime agent is a self-improving rlm agent for long autonomous coding work. it revises its own policy after each run. no cold starts.
trust an agent that edits itself?
github.com/PrimeIntelle...
prime agent is a self-improving rlm agent for long autonomous coding work. it revises its own policy after each run. no cold starts.
trust an agent that edits itself?
github.com/PrimeIntelle...
stop letting your agent grep a monorepo and hope.
code graph rag builds a knowledge graph of your codebase so agents query real structure, not string matches.
how big is your repo?
github.com/vitali87/cod...
code graph rag builds a knowledge graph of your codebase so agents query real structure, not string matches.
how big is your repo?
github.com/vitali87/cod...
August 15, 2026 at 12:00 PM
stop letting your agent grep a monorepo and hope.
code graph rag builds a knowledge graph of your codebase so agents query real structure, not string matches.
how big is your repo?
github.com/vitali87/cod...
code graph rag builds a knowledge graph of your codebase so agents query real structure, not string matches.
how big is your repo?
github.com/vitali87/cod...
durable objects, but running on your own boxes.
celld is a self-hosted distributed durable objects runtime in rust, from the deno team. stateful serverless without renting someone else's edge.
where does your agent state actually live?
github.com/denoland/celld
celld is a self-hosted distributed durable objects runtime in rust, from the deno team. stateful serverless without renting someone else's edge.
where does your agent state actually live?
github.com/denoland/celld
August 15, 2026 at 9:15 AM
durable objects, but running on your own boxes.
celld is a self-hosted distributed durable objects runtime in rust, from the deno team. stateful serverless without renting someone else's edge.
where does your agent state actually live?
github.com/denoland/celld
celld is a self-hosted distributed durable objects runtime in rust, from the deno team. stateful serverless without renting someone else's edge.
where does your agent state actually live?
github.com/denoland/celld
agents that still remember what they were doing an hour later.
loopx holds loop state outside the agent, so progress survives resets and tool handoffs.
minute five or hour two?
github.com/huangruiteng...
loopx holds loop state outside the agent, so progress survives resets and tool handoffs.
minute five or hour two?
github.com/huangruiteng...
August 15, 2026 at 8:30 AM
agents that still remember what they were doing an hour later.
loopx holds loop state outside the agent, so progress survives resets and tool handoffs.
minute five or hour two?
github.com/huangruiteng...
loopx holds loop state outside the agent, so progress survives resets and tool handoffs.
minute five or hour two?
github.com/huangruiteng...
one photo in, a real 3d model out, on your own gpu.
modly turns images into 3d meshes with local ai. no upload, no queue, no per render credits.
what would you scan first?
github.com/lightningpix...
modly turns images into 3d meshes with local ai. no upload, no queue, no per render credits.
what would you scan first?
github.com/lightningpix...
August 14, 2026 at 7:00 PM
one photo in, a real 3d model out, on your own gpu.
modly turns images into 3d meshes with local ai. no upload, no queue, no per render credits.
what would you scan first?
github.com/lightningpix...
modly turns images into 3d meshes with local ai. no upload, no queue, no per render credits.
what would you scan first?
github.com/lightningpix...
one endpoint, every model, zero rewrites.
switchyard routes llm traffic across providers while keeping native openai and anthropic api shapes.
how many providers are you wired to?
github.com/NVIDIA-NeMo/...
switchyard routes llm traffic across providers while keeping native openai and anthropic api shapes.
how many providers are you wired to?
github.com/NVIDIA-NeMo/...
August 14, 2026 at 5:30 PM
one endpoint, every model, zero rewrites.
switchyard routes llm traffic across providers while keeping native openai and anthropic api shapes.
how many providers are you wired to?
github.com/NVIDIA-NeMo/...
switchyard routes llm traffic across providers while keeping native openai and anthropic api shapes.
how many providers are you wired to?
github.com/NVIDIA-NeMo/...
your agent draws diagrams that do not look like ai made them.
29 editorial diagram types for claude code. self-contained html and svg, no mermaid slop.
kept any agent diagram lately?
github.com/cathrynlaver...
29 editorial diagram types for claude code. self-contained html and svg, no mermaid slop.
kept any agent diagram lately?
github.com/cathrynlaver...
August 14, 2026 at 3:30 PM
your agent draws diagrams that do not look like ai made them.
29 editorial diagram types for claude code. self-contained html and svg, no mermaid slop.
kept any agent diagram lately?
github.com/cathrynlaver...
29 editorial diagram types for claude code. self-contained html and svg, no mermaid slop.
kept any agent diagram lately?
github.com/cathrynlaver...
one runtime instead of nine subscriptions.
kun is a local-first agent workspace for coding, writing, research and automation. your files stay on your disk.
consolidate, or keep the specialised tools?
github.com/KunAgent/Kun
kun is a local-first agent workspace for coding, writing, research and automation. your files stay on your disk.
consolidate, or keep the specialised tools?
github.com/KunAgent/Kun
August 14, 2026 at 12:01 PM
one runtime instead of nine subscriptions.
kun is a local-first agent workspace for coding, writing, research and automation. your files stay on your disk.
consolidate, or keep the specialised tools?
github.com/KunAgent/Kun
kun is a local-first agent workspace for coding, writing, research and automation. your files stay on your disk.
consolidate, or keep the specialised tools?
github.com/KunAgent/Kun
one GPU is never the bottleneck, the scheduler is.
dynamo serves inference across a whole fleet. routing, batching, KV cache reuse at datacenter scale.
self-host inference, or pay the API bill?
github.com/ai-dynamo/dy...
dynamo serves inference across a whole fleet. routing, batching, KV cache reuse at datacenter scale.
self-host inference, or pay the API bill?
github.com/ai-dynamo/dy...
August 14, 2026 at 8:30 AM
one GPU is never the bottleneck, the scheduler is.
dynamo serves inference across a whole fleet. routing, batching, KV cache reuse at datacenter scale.
self-host inference, or pay the API bill?
github.com/ai-dynamo/dy...
dynamo serves inference across a whole fleet. routing, batching, KV cache reuse at datacenter scale.
self-host inference, or pay the API bill?
github.com/ai-dynamo/dy...
vector search forgets how things connect.
falkordb is a graph db on sparse matrix math. relationship queries in milliseconds. built for agent memory.
is graph the missing half of RAG?
github.com/FalkorDB/Fal...
falkordb is a graph db on sparse matrix math. relationship queries in milliseconds. built for agent memory.
is graph the missing half of RAG?
github.com/FalkorDB/Fal...
August 13, 2026 at 7:00 PM
vector search forgets how things connect.
falkordb is a graph db on sparse matrix math. relationship queries in milliseconds. built for agent memory.
is graph the missing half of RAG?
github.com/FalkorDB/Fal...
falkordb is a graph db on sparse matrix math. relationship queries in milliseconds. built for agent memory.
is graph the missing half of RAG?
github.com/FalkorDB/Fal...
your agent can now read your whatsapp.
an mcp server that hands your chat history to any agent. search years of threads, draft replies. runs locally.
useful, or a step too far?
github.com/lharries/wha...
an mcp server that hands your chat history to any agent. search years of threads, draft replies. runs locally.
useful, or a step too far?
github.com/lharries/wha...
August 13, 2026 at 5:30 PM
your agent can now read your whatsapp.
an mcp server that hands your chat history to any agent. search years of threads, draft replies. runs locally.
useful, or a step too far?
github.com/lharries/wha...
an mcp server that hands your chat history to any agent. search years of threads, draft replies. runs locally.
useful, or a step too far?
github.com/lharries/wha...
a kanban board where the cards move themselves.
kandev runs coding agents as your team. drag a ticket to in-progress, an agent picks it up and opens the PR.
let an agent pull its own tickets?
github.com/kdlbs/kandev
kandev runs coding agents as your team. drag a ticket to in-progress, an agent picks it up and opens the PR.
let an agent pull its own tickets?
github.com/kdlbs/kandev
August 13, 2026 at 3:30 PM
a kanban board where the cards move themselves.
kandev runs coding agents as your team. drag a ticket to in-progress, an agent picks it up and opens the PR.
let an agent pull its own tickets?
github.com/kdlbs/kandev
kandev runs coding agents as your team. drag a ticket to in-progress, an agent picks it up and opens the PR.
let an agent pull its own tickets?
github.com/kdlbs/kandev