AI & Tech Developments - Feb 18

  • 03:23 β€” OpenClaw 2026.2.17 released: Sonnet 4.6, 1M context beta, subagent spawning from chat, iOS share extension, Slack streaming, Telegram inline buttons β€” described as “chunkiest changelog yet.” @openclaw
  • 12:00 β€” OpenClaw update recap: Kimi K2.5 now free, agents run without paid APIs, skills saved and reused, everything faster. @RoundtableSpace
  • 02:46 β€” New open-source tool coordinates AI agents through shared files and a dashboard: github.com/getclawe/clawe. @tom_doerr
  • 13:35 β€” Tip: build a personal skills library for your OpenClaw agent by running a weekly cron to surface repetitive tasks and auto-generate skill markdown files. @austin_hurwitz
  • 12:38 β€” Becoming a “10X engineer” now means maintaining a SKILLS.md file for your AI agent. @mdancho84
  • 12:12 β€” Developer ran an AI self-review loop every 4 hours for 30 days using SOUL.md (personality) and MEMORY.md (persistent memory) files, reporting notable self-correction improvements. @kaostyl
  • 12:39 β€” Developer replaced all 15+ n8n automation workflows by exporting them as JSON and having AI rebuild them natively, declaring “n8n is dead.” @kaostyl
  • 00:28 β€” Tip: override Claude Code’s model settings in .claude/settings.json to use claude-sonnet-4-6-1m for both haiku and sonnet slots, unlocking the 1M context window. @melvynxdev @cloudxdev
  • 23:12 β€” Developer used Opus 4.6 1M + Sonnet 4.6 1M sub-agents to do the largest technical planning session yet: reviewed full docs and repo, generated a Linear project with 7 epics, 40+ sub-issues, and full dependency tracking. @nummanali
  • 03:19 β€” Levelsio reports running OpenClaw connected to production servers via Telegram after 3 weeks of testing, enabling live feature builds and bug fixes through chat; now extending the approach to other sites on isolated VPSes. @levelsio
  • 18:52 β€” Developer argues AI agent swarms should move to Discord for voice: morning briefings while driving, live verbal debugging, swarm group discussions. @jumperz
  • 17:26 β€” Discord highlighted as a free, complete infrastructure layer for AI agents: channels, threads, forums, voice, APIs, security β€” stop building custom infra from scratch. @jumperz
  • 11:30 β€” Warning: scaling to 9+ AI agents causes silent failures β€” stale context, dropped tasks, invisible errors. Coordination overhead becomes critical. @jumperz
  • 19:30 β€” Ex-Snapchat engineers launch @layers, an AI CMO that reads your GitHub repo and runs the full marketing stack autonomously. @hasantoxr @yoheinakajima
  • 17:16 β€” Fortune profiles 22-year-old dropout Adavia Davis making $40K–$60K/month from 5 fully AI-generated faceless YouTube channels at 85–89% profit margins; Fiverr separately reports 488% surge in demand for faceless YouTube creators. @knoxtwts
  • 05:02 β€” Developer stack for agentic coding: Ghostty + tmux + Neovim + OpenCode (GPT-5.3-Codex). @willccbb
  • 03:08 β€” Key prompt pattern for overnight GPT-5.3-Codex builds: define a strict build contract with deliverables, tests, reports, and stop conditions before launching. @morganlinton
  • 09:55 β€” Simple persistent memory trick for Claude: add a LEARNING.md file to your project and reference it in CLAUDE.md for the agent to read and update its own lessons. @im_bulent
  • 04:59 β€” Best practice: write general coding patterns in ~/.claude global CLAUDE.md to solve in-context learning issues across all codebases and languages. @KingBootoshi
  • 18:45 β€” Tip: give OpenClaw a copy of your Chrome profile so it’s pre-logged-into all your services for browser automation (use with caution). @mernit
  • 18:02 β€” Warning: blindly building through chat with OpenClaw leads to architecture drift and misunderstood requirements β€” treat it like a human employee and align before executing. @chrysb
  • 10:04 β€” Developer uses OpenClaw daily: 6am weather + newsletter digest, Matt Levine summaries, and more automated morning briefings. @frankdilo
  • 11:07 β€” Prompt idea: upload credit card statements to OpenClaw and ask it to identify subscriptions replaceable with open-source alternatives or DIY builds. @johann_sath
  • 19:00 β€” Developer built a Claude Code pipeline to scrape LinkedIn post engagers and auto-send cold emails to ideal customer profiles. @codyschneiderxx
  • 05:35 β€” Best practice: onboard OpenClaw like an employee β€” dedicated email, scoped API keys, isolated machine, written SOPs, and regular reviews. @johann_sath
  • 00:52 β€” Two business opportunities highlighted: selling OpenClaw/agent infra B2B, and running fully automated AI UGC affiliate video channels. @mike_revenue
  • 15:55 β€” AI-as-a-service framed as the new consulting: identify workflows, set up personalized agents, profit. @0xNairolf
  • 16:37 β€” Businesses quietly going all-in on Claude + SEO framed as analogous to buying Bitcoin at $100 β€” a stealth advantage most aren’t taking seriously yet. @bloggersarvesh
  • 16:00 β€” YC founder casually mentions replacing their entire sales team with 14 AI agents handling outbound, onboarding, and support β€” shown live on a dashboard. @shiftj
  • 16:17 β€” Developer built a to-do agent that nudges until tasks get done, does research, and pings across Discord, email, and Mac notifications β€” built with the new agents SDK. @burcs
  • 11:36 β€” Comprehensive AI tools roundup: Claude, Gemini, GPT, Kimi for LLMs; Claude Code, Cursor, OpenCode, Lovable for coding; Manus/Claude for computer-use; NanoBanana Pro, GPT-image, Midjourney for images; and more. @eptwts
  • 18:41 β€” Anthropic applied context engineering to web search: Sonnet 4.6 now writes and executes code to filter search results before they hit the context window. @himanshustwts
  • 19:00 β€” Claude’s web search and fetch tools now write and execute code to filter results before reaching the context window β€” 13% higher accuracy on BrowseComp with 32% fewer input tokens. @alexalbert__
  • 12:31 β€” Demo: AI method to convert a 2D floor plan into a 3D walkthrough tour using Freepik Spaces. @techhalla
  • 19:38 β€” Pointer to a masterclass video on Claude Code from its creator; Peter Steinberger (PSPDFKit founder, OpenClaw creator) joins OpenAI. @aakashgupta
  • 14:44 β€” Tip: prompt Claude Code with your idea and ask it to “ask clarifying questions until you know what I want to build” for better results. @alexhillman
  • 03:07 β€” New OpenClaw skill /moveto lets you hand off tasks between interfaces (Discord, web UI, email) β€” enables seamless context switching across workflows. @alexhillman
  • 16:48 β€” Mac productivity tip: exclude code project directories from Spotlight indexing to prevent massive memory drain from indexing hundreds of thousands of small files. @billyjhowell
  • 15:21 β€” Fun hook: add WarCraft 3 sounds to Claude hooks for audio alerts when tasks finish or need permission. @JorgeCastilloPr
  • 11:27 β€” Best practice for reviewing agent work: have the agent generate a structured report, send it to an e-ink tablet, review over a martini, request changes via Discord. @__morse
  • 01:07 β€” Developer switches from Claude Code to Codex after 6 months: Codex described as “autistic genius” that gets things done quietly vs. Claude Code’s speed being “an illusion.” @Hesamation
  • 21:29 β€” Claude Opus 4.6 wrote vulnerable Solidity code leading to a $1.78M smart contract exploit: cbETH price set to $1.12 instead of ~$2,200 β€” possibly the first AI vibe-coded DeFi hack. @pashov
  • 18:59 β€” Tip: use free Kimi K2.5 via NVIDIA’s build platform as a cost-free alternative to OpenClaw API (comparable to Opus 4.5); DHH separately reports Kimi K2.5 on OpenCode Zen fixed a bug in 21s vs Claude’s 3 minutes, at near-zero cost β€” $10 left after 3M tokens. @KSimback @dhh
  • 02:27 β€” Building with OpenClaw day 17–23: $5K Mac Studio, Valentine’s AI relationship chat, perfected coding agent swarm spawn setup, Perplexity + Brave API integration. @elvissun
  • 21:42 β€” Demo: build your own AI-native self-hosted Notion/Evernote with full Claude Code access. @Shpigford
  • 15:17 β€” Design prompt tip: tell the LLM to break its default rule-following tendencies to get more original, unconventional creative output. @dayonefoundry
  • 05:29 β€” Roundup of 21 real OpenClaw workflows by Matthew Berman: personal CRM from Gmail/calendar, knowledge base, and more practical no-fluff automations. @code_rams
  • 01:36 β€” Apple’s AI strategy: do nothing, avoid $100B capex, pick the winning model (Gemini), and build on top β€” framed as genius free cash flow play. @cryptopunk7213
  • 18:39 β€” Firecrawl launches Browser Sandbox: zero-config, fully managed secure browser environments for AI agents, compatible with Claude Code and Codex. @firecrawl
  • 20:01 β€” AI 3D consistency demo: generation quality now “actually crazy” β€” possibly making traditional 3D learning obsolete. @hewarsaber
  • 04:41 β€” Peter Steinberger (PSPDFKit bootstrapper, OpenClaw creator) joins OpenAI, leaving Vienna for San Francisco. @aakashgupta
  • 05:48 β€” Google Research finding: simply pasting a prompt twice yields 47 wins out of 70 benchmark tests across 7 models with zero losses β€” described as damning for current LLM architecture. @aakashgupta
  • 20:18 β€” Developer connected Claude Sonnet 4.6 to Webflow MCP, had it audit and auto-fix 15–20 SEO issues across sites in 5 minutes. @KurtisHanni
  • 22:26 β€” Next.js devtools MCP flagged as a killer tool for AI-assisted Next.js development. @ryancarson
  • 21:06 β€” Phantom MCP Server launched: agents can swap, sign, and manage addresses across all of Phantom’s supported chains β€” compatible with Claude, OpenClaw, and any MCP client. @phantom
  • 03:57 β€” Developer experiments with running OpenClaw overnight on a task queue: finds naive approaches either do one task at a time or overwhelm a VPS β€” exploring coordination solutions. @jeanslofi
  • 09:41 β€” Open-source PentAGI project: multi-agent AI red team where separate agents handle recon, scanning, exploitation, and report writing with zero human input β€” agents coordinate and adapt to findings. @chiefofautism
  • 23:19 β€” Three-stage evolution of AI agent context management: Stage 1 is tool bloat (most production agents still here), Stage 2+ moves toward selective context and smarter retrieval. @koylanai
  • 14:55 β€” Garry Tan frames the new founder question: how do you make AI agents discover and choose your product? Old SEO/marketing tricks won’t work. @garrytan
  • 01:56 β€” Observer notes the largest knowledge gap in tech history: ~1 million people using Codex/Claude vs. the rest of humanity β€” and the gap is growing. @draecomino
  • 03:17 β€” AI-powered school reportedly outperforms top 0.1% of schools globally: average student equivalent to a valedictorian at a top 15% US private school; mathematics scores break standard scaling. @cryptopunk7213
  • 09:10 β€” OpenClaw hacks thread: use forum posts for long-term task sessions, build UI components via Discord, feed goals/context/socials for richer agent output. @0xSero
  • 18:27 β€” OpenClaw’s Larry skill goes mainstream: competitor research, content generation based on what’s working, view/revenue tracking β€” two years of manual marketing replaced. @oliverhenry
  • 01:51 β€” OpenClaw use cases roundup: targeted morning briefings on news, stocks, futures, and AI developments β€” highly customizable and addictive. @godwinbabu
  • 23:00 β€” Email agents highlighted as underrated: AI monitors inbox, responds to cold email replies, and drives real-world GTM outcomes autonomously. @codyschneiderxx
  • 20:12 β€” Camofox v1.2: browser-for-agents tool now launches on demand and shuts down after 5 min idle, dropping memory to ~40MB β€” built for $5 VPS and Raspberry Pi stacks. @pradeep24
  • 19:06 β€” Developer built a $0.07/post cross-platform publisher hitting 5 platforms simultaneously. @mdnlabs
  • 13:41 β€” Claude Code Lamp: physical BLE lamp integration with Claude Code hooks gives visual feedback when the agent needs input or is working. @bobek_balinek
  • 10:52 β€” Best practice for AI coding agents writing tests: “Don’t write tests for what the type system already guarantees” β€” cuts an entire class of useless tests. @mattpocockuk
  • 09:03 β€” Voicebox launched: open-source local voice cloning from a few seconds of audio, powered by Qwen3-TTS β€” no cloud, no subscription, no data leaves device. @JafarNajafov @hasantoxr
  • 17:29 β€” SuperCmd launched: open-source Raycast alternative with ElevenLabs voice dictation, compatible with all existing Raycast extensions β€” zero migration. @nullbytes00
  • 15:42 β€” Local macOS voice control app found running Parakeet V3 fully on-device. @victormustar
  • 07:10 β€” Claude Sonnet 4.6 framed as AGI signal: developers preferred it over the previous frontier Opus model 59% of the time in early testing. @alex_prompter
  • 20:29 β€” Agent skill “Visual Explainer” built to reduce cognitive debt: converts complex topics into rich HTML pages using a CSS pattern library for consistent output. @nicopreme
  • 13:42 β€” Web-based System Design Simulator: drag-and-drop architecture components with real-time traffic, failure, latency, and scaling simulation. @Amank1412
  • 18:05 β€” Liveline: open-source real-time chart library for live markets β€” animated, with time-window controls, orderbook components, and particle bursts. @benjitaylor
  • 09:35 β€” French dev notes Claude now proactively asks clarifying questions before answering to better frame the response β€” sees it as a new level of capability. @eliasghostw
  • 14:13 β€” Viral claim: watching a Mac complete a 4-hour workday in 90 seconds via AI automation. @bigaiguy
  • 11:36 β€” eCom veteran amazed that sending an Amazon product link to OpenClaw agent can instantly generate creative assets β€” a workflow that previously cost tens of thousands of dollars. @GanimCorey
  • 00:53 β€” Sigil builds a platform where AI agents are funded to pay for their own hosting, compute, and investments β€” agents that earn enough replicate by buying new servers and copying files. @SCHIZO_FREQ
  • 05:10 β€” ASCII wireframe prototyping tool launched: mockdown.design. @bbssppllvv
  • 20:49 β€” Concept pitch: TikTok-style doom-scroll interface for reviewing and accepting background AI agent PR suggestions on your codebase. @almonk
  • 19:04 β€” Advanced SEO content workflow: scrape top 20 SERPs, build brand-specific AI project, train on existing quality content, detailed outlines matching search intent. @Nick_zv_
  • 17:00 β€” Reality check tweet: “Are you brave enough to admit you don’t have a use case for OpenClaw?” @zeroxjackson
  • 06:52 β€” Curated list of public OpenClaw skills now available via Vercel Skills directory at skills.sh. @cloudxdev

πŸ“± Source Tweets

OpenClaw 2026.2.17 just dropped Sonnet 4.6 1M context beta spawn subagents from chat iOS share extension Slack native streaming styled inline buttons on Telegram chunkiest changelog yet. upgrade or get left behind https://github.com/openclaw/openclaw/releases/tag/v2026.2.17…

β€” @openclaw

Coordinates AI agents through shared files and a dashboard https://github.com/getclawe/clawe

β€” @tom_doerr

building your own skills library is the fastest way to supercharge your OpenClaw agent run a cron each week to review any task you did more than once have your agent surface and create skill markdown files for anything you plan to do again you've now trained your agent on your

β€” @austin_hurwitz

I asked my AI to review itself. Every 4 hours. Here's what it found after 30 days. Most people use AI as a chatbot. I gave mine a personality, a memory, and a self-correction loop. http://SOUL.md β€” defines WHO my AI is http://MEMORY.md β€” persistent memory

β€” @kaostyl

n8n is dead. I fed my AI all my workflow JSONs and it replaced every single one. I ran n8n for over a year. 15+ workflows on a Hetzner VPS. HTTP nodes, webhooks, JSON parsing, conditional branches, error handling. Then I did something stupid: I exported every workflow as JSON,

β€” @kaostyl

PRO TIP: Unlock the biggest update for your Claude Code now: 1. Go to .claude/settings.json and add this: "env": { "ANTHROPIC_DEFAULT_HAIKU_MODEL": "claude-sonnet-4-6-1m", "ANTHROPIC_DEFAULT_SONNET_MODEL": "claude-sonnet-4-6-1m" } 2. Use /model to switch to Sonnet Enjoy

β€” @melvynxdev

Update your ~/.claude/settings.json for your subagents to use sonnet 4.6 instead of haiku or sonnet 4.5

β€” @cloudxdev

A dream I had finally came true: I can now chat directly with my sites to build any feature or fix any bug just via Telegram I've been playing with OpenClaw for 3 weeks now and it's great but I was always too scared to run it on any production server And I was right a bit as

β€” @levelsio

if you haven't pivoted to discord you prolly should voice calls gonna be insane.. imagine morning briefings from your swarm while you're driving, or live debugging where you just ask an agent what went wrong and it explains out loud.. or simply your whole swarm discussing

β€” @jumperz

i've been running exactly 9 agents and honestly, when you start scaling alot of things fail in the background, you will have no idea what happened and the more agents you add, the worse it gets your system looks up, everything is running, but context is stale, some tasks get

β€” @jumperz

Holy shit... 3 ex-Snapchat engineers just built an AI CMO that reads your GitHub and runs your entire marketing stack. It's called @layers and it quietly kills the need for marketing agencies, freelancers, and $10K/mo CMOs. Here's everything you need to know:

β€” @hasantoxr

fortune just profiled a 22 year old dropout named adavia davis clearing $40,000-$60,000/mo from faceless youtube channels. all ai generated. every single one. his operating cost is $6,500/mo. profit margin sits between 85-89%. he runs 5 channels. history. minecraft. animals.

β€” @knoxtwts

turns out this is just ghostty + tmux + neovim + opencode (5.3-codex)

β€” @willccbb

I'm starting to get into a better groove with GPT-5.3-Codex for overnight builds. This is the key sentence that's working for me right now: "define a strict build contract (deliverables, tests, reports, and stop conditions) so it's clearly an overnight-sized feature before

β€” @morganlinton

Tout le monde cherche un "Longterm Memory MCP" sophistiquΓ© pour que Claude se souvienne de ses apprentissages. La solution ? Un fichier http://LEARNING.md dans ton projet. C'est tout. En gros tu ajoutes Γ§a dans ton http://CLAUDE.md : "Lis http://LEARNING.md

β€” @im_bulent

PSA to everyone building browser agents with @openclaw give it a copy of your chrome profile now it's logged into everything in your browser (use with caution)

β€” @mernit

blindly telling your @openclaw what to build through chat works until it doesn't. eventually, half-baked implementations compound through misunderstood requirements and architecture drift, with no paper trail. just like a human employee, alignment before execution becomes more

β€” @chrysb

I've been using @OpenClaw daily for weeks now. Here's what I actually do with it. Morning brief Every day at 6am, it sends me weather + a digest of newsletters from my RSS reader (Feedbin). Summarizes Matt Levine's Money Stuff so I get the key points in 2 min. Also

β€” @frankdilo

upload your credit card statements to OpenClaw "what subscriptions can we replace with open source or build ourselves?"

β€” @johann_sath

I see hundreds of linkedin posts every day that my perfect ICP is commenting on here's how I made Claude Code find them and cold email them for me hey claude, build this: I'm going to give you a LinkedIn post and I want you to extract the engagers linkedin URLs I'm going to

β€” @codyschneiderxx

onboarding OpenClaw is like onboarding an employee: 1. its own email (not yours) 2. its own API keys (not your personal ones) 3. its own machine 4. scoped permissions 5. written SOPs 6. regular reviews

β€” @johann_sath

There are only two businesses I'd recommend anyone who is just starting do rn: 1. Sell OpenClaw installations and agent management infra B2B 2. Make AI ugc vids and drive millions of views to affiliate products on organic (this can be fully automated btw) no noise no bs. these

β€” @mike_revenue

starting to see ai-as-a-service as an insane business - help people/companies find the best workflows to optimize - set up personalized ai agents for them - profit feels like the new consulting

β€” @0xNairolf

This started as a flat 2D floor plan. and I'm sharing the exact AI method I used to turn it into a 3D tour inside freepik Spaces

β€” @techhalla

Instead of scrolling, watch this masterclass on Claude Code from the creator

β€” @aakashgupta

Literally set up Claude Code and post these words into it with the additional "ask me clarifying questions until you know what I want to build and walk me through the setup step by step." The beautiful thing about these systems is that if you can describe what you want you don't

β€” @alexhillman

Mac users: I found a fix for a massive memory drain Each time you boot up your MacBook, Spotlight tries to index every single tiny file. If you code a lot, this could be tens or hundreds of thousands of files Do yourself a GIGANTIC favor and go to Settings > Spotlight >

β€” @billyjhowell

How to finally become the 10x engineer: Add WarCraft 3 sounds to Claude hooks to get alerts when it finishes a task or needs permission.

β€” @JorgeCastilloPr

Kimi K2.5 on @opencode Zen is hilariously cheap. I bought $20 worth of tokens two weeks ago, and I still have $10.89 left! After 3M tokens! If there's a bubble in AI, it's pricing a million tokens at $25 (and beyond).

β€” @dhh

Kevin: "I just raced Claude and Kimi K2.5 against that bug that Ryan was talking about. K2.5 fixed it in 21s. Claude took just over a minute to make the plan, then about 2 minutes to execute on it. Both had the same fix, though." (K2.5 is now my main driver. Opus just backup.)

β€” @dhh

This paper is quietly one of the most damning findings about current LLM architecture. Google Research tested 7 models across 7 benchmarks. The intervention was embarrassingly simple: paste the prompt twice. The result: 47 wins out of 70 tests, zero losses.

β€” @aakashgupta

I connected Claude Sonnet 4.6 to the Webflow MCP and asked it what issues my sites had. It pointed out 15-20 SEO issues and the fixes, then asked if I wanted it to fix them. Said yes, and 5 minutes later, they were all fixed.

β€” @KurtisHanni

Eu queria que o Openclaw trabalhasse enquanto eu durmo, entΓ£o meu primeiro teste foi: - vou criar uma lista de tarefas e vou deixar o agente trabalhar mas isso causava alguns problemas: 1. ou o agente sΓ³ fazia 1 task por vez 2. ou fazia todas de uma vez e derrubava meu VPS

β€” @jeanslofi

I built this. Costs me $0.07/post. It will hit 5 platforms at once.

β€” @mdnlabs

someone built an entire AI RED TEAM - multiple agents that coordinate HACKING ATTACKS together, ZERO human input PentAGI, open source, one agent does recon, another scans, another exploits, another writes the report. they talk to each other and adapt based on what they find

β€” @chiefofautism

Introducing Claude Code Lamp : Get visual feedback whenever your code agent needs input or is busy coding away. It's just a nifty integration of @claudeai Code hooks and a BLE lamp. Full source code in the replies

β€” @bobek_balinek

I think we are getting AGI before 2027. Anthropic just dropped Claude Sonnet 4.6 and its a monster. Here's what nobody is talking about: Developers who tested it early preferred Sonnet 4.6 over the previous frontier Opus model 59% of the time.

β€” @alex_prompter

I finally found the app I wanted for voice control on macOS. 100% local, running Parakeet V3 extremely fast.

β€” @victormustar

Je demande un truc Γ  Claude ce matin il me sort un questionnaire parce qu'il veut Γͺtre sΓ»r de bien cadrer la question et de me faire la meilleure rΓ©ponse.. mais ptn c'est quoi ce juice ?

β€” @eliasghostw

Someone built a web-based System Design Simulator, where you drag & drop architecture components and actually simulate traffic, failures, latency, and scaling in real time, System design just got way more interactive.

β€” @Amank1412

We're watching a three-stage evolution in how AI agents manage context. Stage 1: Tool bloat. Give the agent every tool, every connector, every context; tool descriptions, retrieved docs, conversation history. Performance degrades. Most production agents are still here. Stage 2:

β€” @koylanai

One of the most important questions for founders is: How do I make sure agents know about my product and service and choose it? All the old tricks won't work. People who figure this out will win big

β€” @garrytan

I've followed tech for 25 years and I've never felt a larger gap between the ~1 million people using Codex/Claude and the rest of humanity.

β€” @draecomino

this school used AI to teach kids and outperformed the top 0.1% of schools globally, the stats are insane

β€” @cryptopunk7213

New Openclaw hacks. 1. Task management: - Use forum for long term life tasks, each post under forum is it's own session - Use the new feature to allow openclaw to build you UI components in discord - Feed it your problems, ambitions, goals, context, socials.

β€” @0xSero

Larry has gone full mainstream. It took me two years of trying to finally automate my marketing Thanks to @openclaw I finally have everything I need. The full skill > researches competitors > generates content based on what is working for them > tracks views and paying

β€” @oliverhenry

Been playing with #OpenClaw since its ClawdBot days, right as it was catching fire. Some of the ways I am using it: - Targeted morning briefings on news, stocks & futures, AI developments

β€” @godwinbabu

agents that just run email for you are the most underrated thing in your tool kit like the fact that I can cold email people and have an agent monitor the inbox constantly and respond and it just makes real world outcomes happen for go to market

β€” @codyschneiderxx

Created an agent skill called "Visual Explainer" + set of complementary slash commands aimed to reduce my cognitive debt so the agent can explain complex things as rich HTML pages.

β€” @nicopreme

camofox v1.2 is out -- browser launches on-demand, shuts down after 5 minutes idle, and relaunches instantly when needed. idle mem now ~40MB for resource-constrained agent environments like $5 VPS and Raspberry Pi stacks.

β€” @pradeep24

Holy shit... someone just built the "Ollama for voice" and it's completely opensource. It's called Voicebox and it clones any voice from a few seconds of audio. No cloud. No subscription. No voice data ever leaving your machine. The model behind it is Qwen3-TTS from Alibaba.

β€” @JafarNajafov

BREAKING: The "Ollama for voice cloning" just dropped. It's called Voicebox and it clones any voice from just a few seconds of audio entirely on your machine.

β€” @hasantoxr

We just shipped SuperCmd - The ultimate open-source alternative to @raycast The wildest part? Every Raycast extension already works with it. Zero migration. Just install and go.

β€” @nullbytes00

Important instruction for AI Coding agents writing tests: "Don't write tests for what the type system already guarantees." Helps hint against an entire class of shit tests.

β€” @mattpocockuk

Are you brave enough to admit you don't have a use case for open claw?

β€” @zeroxjackson

RIP typing. I just watched my Mac finish a 4-hour workday in 90 seconds. This is absurd.

β€” @bigaiguy

Liveline: a Polymarket-inspired real-time chart library for live markets (e.g. 15-minute BTC windows) or any streaming data. Fully animated, with built-in time-window controls, orderbook components, particle bursts, and more.

β€” @benjitaylor

here are all the AI tools you should get up to speed with right now...

β€” @eptwts

Opus 4.6 1M Main Sonnet 4.6 1M Sub Agents Just did the largest technical planning work ever...

β€” @nummanali

honestly...the businesses quietly going all-in on Claude + SEO right now are going to be filthy rich this year.

β€” @bloggersarvesh

finally built the agent i've been needing for a minute now a todo list that nudges you until stuff actually gets done.

β€” @burcs

discord gave us all of this for free and people are still building custom infrastructure from scratch.

β€” @jumperz

anthropic did CONTEXT ENGINEERING on WEB SEARCH

β€” @himanshustwts

Sigil is truly a genius for this one He made a platform that lets you fund an AI agent with enough money to pay for its own server hosting, compute, and investments

β€” @SCHIZO_FREQ

At a YC dinner last week, a founder casually mentioned they're not hiring sales anymore. "14 AI agents."

β€” @shiftj

Underrated dev upgrade from today's launch: Claude's web search and fetch tools now write and execute code to filter results before they reach the context window.

β€” @alexalbert__

New skill I made this weekend that has already changed the game for me: /moveto

β€” @alexhillman

THE NEW OPENCLAW UPDATE IS GOATED - Kimi K2.5 is now free - Agents run without paid APIs

β€” @RoundtableSpace

The next-devtools mcp is killer. If you're building in @nextjs check it out.

β€” @ryancarson

Introducing the Phantom MCP Server Agents can swap, sign, and manage addresses across all of Phantom's supported chains.

β€” @phantom

I spent 9 years in eCom. So when I say this tech is mind blowing, I mean it.

β€” @GanimCorey

every single problem i see people have with coding agents can be solved by informing them

β€” @KingBootoshi

this team convinced me to invest by explaining how devs hate marketing so much, they don't even want a marketing app

β€” @yoheinakajima

ok hear me out - TikTok but for accepting PR suggestions from background agents running over your codebases

β€” @almonk

Quick ASCII wireframe prototyping tool http://mockdown.design

β€” @bbssppllvv

How you write SEO content: "Jarvis, write me a blog post" How I write SEO content: β†’ Scrape top 20 SERPs...

β€” @Nick_zv_

OpenClaw 2026.2.17 released

β€” @openclaw

All Public Skills, Bookmark it β€” skills.sh/?q=CloudAI-X

β€” @cloudxdev