Coolify
Self-hosted Heroku and Vercel alternative
Dev tools & infraApache-2.0★ 65.5kv4.6.0
AI news
PocketBase
Open-source backend in one file: SQLite, auth, files, realtime and an admin dashboard
Dev tools & infraMIT★ 61.6kv0.40.6
Hono
Web framework for any JavaScript runtime
Dev tools & infraMIT★ 32.8kv4.13.16
Bun
All-in-one JavaScript runtime and toolkit
Dev tools & infraMIT★ 96.5kv1.4.4
Ghostty
Fast, feature-rich, native terminal
Dev tools & infraMIT★ 63.3kv1.4.0
Zed
The editor for what's next
Dev tools & infraGPL-3.0★ 92.6kv1.26.2
GPT-SoVITS
Few-shot voice cloning from one minute of audio
Local AI & modelsMIT★ 64.6k
TypingMind
A better UI for your own AI keys
Local AI & models
Cherry Studio
An AI workstation for every model you use
Local AI & modelsAGPL-3.0★ 53.2kv2.1.10
LobeHub
A workspace for organizing and running teams of AI agents
Agents & coding★ 83.7kv2.3.0
Open WebUI
Self-hosted AI interface
Local AI & models★ 157.6kv0.12.5
Exo
Cluster your Macs to run models too big for one
Local AI & modelsApache-2.0★ 48.2kv1.0.71
Ollama
Get up and running with local models
Local AI & modelsMIT★ 185.7kv0.40.13
llama.cpp
LLM inference in pure C/C++
Local AI & modelsMIT★ 133.2kv0.8.0
Tabby
Self-hosted AI coding assistant
Agents & codingApache-2.0★ 33.9kv0.32.0
Val Town
Instant deployments for you and your agents: write TypeScript and ship it to a live URL in moments.
Agents & coding
Onlook
The Cursor for designers
Agents & codingProprietaryv0.2.32
Open Interpreter
A terminal coding agent for open models like Kimi K3 and GLM 5.3
Agents & codingApache-2.0★ 68.8kv0.0.58
Firecrawl
Turn any website into LLM-ready data
Agents & codingAGPL-3.0★ 197.9kv2.11.0
Browser Use
Make websites accessible for AI agents
Agents & codingMIT★ 119.9kv0.13.14
Cline
Open-source coding agent that asks before it edits or runs, with any model
Agents & codingApache-2.0★ 71.2kv4.1.26
Aider
AI pair programming in your terminal
Agents & codingApache-2.0★ 49.9kv0.86.0
Ollama v0.34.0
Ollama v0.34.0 lets Ollama models run inside ChatGPT Desktop, with setup available from the Ollama app on macOS. The release also speeds up structured output on Apple Silicon and supports tool search and response compaction for OpenAI-compatible clients.
OllamaGitHub ReleasesImpeccable v1.4.0
Impeccable's browser extension ext v1.4.0 now snapshots the page and runs the rules as WebAssembly in an offscreen document, so a site's Content Security Policy no longer decides whether the detector works. The new offscreen permission exists only for this, and nothing else about what the extension can see has changed.
ImpeccableGitHub Releases 2 more updatesshadcn/ui v4.21.0
With 4.21.0, running init installs the cn package and writes lib/utils to re-export cn from it, and registry components take cn from that package. Internally, shadcn/ui now uses twMerge from cn rather than tailwind-merge.
shadcn/uiGitHub ReleasesCollecting ideas to improve Midjourney
From the official blog (updates.midjourney.com), published . View original
MidjourneyOfficial blogAlpha Changelog - 9/2/26
From the official blog (updates.midjourney.com), published . View original
MidjourneyOfficial blogBrowser Use can now pay with your credit card using Stripe Link CLI
The agent fills the form and pays. The card is gone after.
Browser UseOfficial blogLatitude v9.11.0
In Latitude typescript-sdk 9.11.0, signals returned by client.signals.list, client.signals.get, client.sessions.listSignals and client.sessions.getSignal include a required scoreEvidence list, with each entry naming the Agent Score dimension it belongs to and the signal's role there. An empty list marks a diagnostic signal.
LatitudeGitHub ReleasesZed v1.18.0
Zed v1.18.0 shows inline values when debugging C and C++ with CodeLLDB or GDB, and adds separate Continue Program and Continue Thread controls for adapters that support single-thread execution. Markdown preview gets new typography and spacing, and File Finder history now includes files opened in previous sessions. On the AI side, models such as Gemini 3.5 Flash-Lite, Grok 4.5 and Grok 4.6 are available, and a broken external agent connection can be reloaded from the Agent Panel without restarting. The release fixes OAuth failing with MCP servers that require non-default scopes.
ZedGitHub ReleasesVal Town Newsletter 27
From the official blog (blog.val.town), published . View original
Val TownOfficial blogFolo v1.13.0
Folo Desktop v1.13.0 can sort unread timelines oldest first, and Linux builds now come with a .deb package next to the AppImage. Copying masked inbox secrets has also been fixed.
FoloGitHub ReleasesPaperclip v2026.831.0
Paperclip v2026.831.0 delivers the company skill library to agents at run time, adds Kimi Code as an adapter, and rebuilds the sandbox callback bridge on a bounded HTTP/2 transport. Operators hosting Paperclip for others can control which settings pages their users see. The release requires Node.js 24.11.0 or newer, and managed installs refuse to run on older versions. Agent credentials that fail verification now return a 401 instead of falling back to the anonymous local user, and the company brand color and per-company attachment size limit fields are removed.
PaperclipGitHub ReleasesLatitude v4.1.0
Latitude typescript-telemetry 4.1.0 adds injectTraceContext, extractTraceContext and withTraceContext to carry the active span across boundaries with no ambient OpenTelemetry context, so an agent in another Cloudflare Durable Object joins the caller's trace. A new createDurableObjectTelemetry flushes at each unit of work, and a failed export is reported rather than failing the turn. Constructing Latitude no longer assumes process.once exists, so it works on runtimes without Node signal handling. The smart filter now promotes ancestors of kept spans so exported traces stay connected.
LatitudeGitHub Releases 1 more updateOllama's transparent pricing
Ollama’s new pricing has no service fees and no 5-hour or weekly limits.
OllamaOfficial blogHow to Dictate into Any Mac App with a Fully Local Speech Model
From the official blog (rapidmlx.com), published . View original
Rapid-MLXOfficial blogLatitude v0.2.0
Latitude hermes-telemetry 0.2.0 records reads and writes of Hermes's built-in memory as gen_ai.memory.* operations and turns itself off when an external memory provider is configured. Tool definitions are exported, so a project's tool rollup can see tools the agent has but never called. Exported content passes through Hermes's own secret redactor by default, which LATITUDE_HERMES_REDACT_SECRETS=0 turns off. The Codex/Responses path no longer drops tool calls, tool results and assistant text, and failed tool calls are no longer recorded as successes.
LatitudeGitHub Releases