Our Mission and Vision is to enable every Malaysian 🇲🇾 by teaching how to wield AI as a tool and a skill.
To promote our movement, please follow our socials 👇
✍️ Threads → https://t.co/c7M8Kp4Lra
▶️ YouTube → https://t.co/u9EB1bsz6r
🎵 TikTok → https://t.co/TR4Ed96JEu
📸 Instagram → https://t.co/YyTC6X9Ug3
👾 Discord → https://t.co/sij8r7z4hv
Share with a friend who needs to see this!
Malaysia doesn't lack AI users 🇲🇾
Malaysia needs more AI builders!
That's the belief we brought to @utm_my Malaysia today
krackeddevs is proud to be speaking and showcasing at their "AI & Business: Turning Ideas into Innovation" Technopreneurship Seminar 👨🏻💻
We're here to show the bright students how to move from AI awareness to AI execution ~ to learn, build, showcase, and ship real digital products, not just talk about them.
To the next generation of Malaysian technopreneurs:
The future belongs to builders.
Don't just use AI. Build with it. 🇲🇾
I just shipped 👉 Pasar API 🛒
You've heard of pasar malam. But pasar API? lol
A Malaysia-first directory for the local stuff that's normally scattered and painful ~ FPX, DuitNow, Billplz, e-invoicing, logistics ~ all in one trusted place.
Why vibe coders + Claude users love it:
→ MCP server = your agent pulls the right Malaysian API mid-build. No tab-hopping.
→ No-auth tier = free gov APIs you call straight from the browser, zero setup.
→ "Exclude, don't fake" curation = no dead or hallucinated endpoints wasting your time.
It kills the worst part of shipping ~ "which API do I use and how do I call it?" ~ and hands you a copy-paste answer.
Malaysia-first, agent-ready.
→ https://t.co/egoOD1mB3u
I just built the KRACKED SQUARE 😎🕹️
a Habbo-style online hangout where everyone in our @KrackedDevs Discord literally appears in the game.
Walk around, chat, throw emotes, jump, and bump into the whole community in a living pixel world.
Built in 2 days with @claudeai Fable 5
Please try it out and give me some feedback guys - https://t.co/YaHsgInak5
Made this with Fable 5 & Opus
Design your own Malaysian map 🇲🇾
Then download it FREE.
For devs and vibe coders building something with a Malaysia map, or designers who just want a clean one to drop into their work.
Export as SVG, GeoJSON, embed, or React component.
👉 https://t.co/EjWVT9z7kR
Introducing mybitduit ~ a @solana terminal that lives in your command line.
Live prices, charts drawn right in the terminal, and self-custody wallets that never touch your keys.
Non-custodial. Made in Malaysia 🇲🇾
Bit + Duit - https://t.co/UBGXyOhmpr
It’s going to be a busy week for us 😁
Kicking it off today ~
We’re speaking & showcasing our products at an Tech & AI Educational Event in Melaka ~ an initiative by the Malaysian Government 
@KrackedDevs x @TalentCorpMsia
Anthropic has launched Claude Corps, a national fellowship program that embeds 1,000 early career AI trained fellows inside nonprofits across the United States for a year, fully funded by Anthropic. 🎓
The goal is to put powerful AI tools to work for mission driven organizations while building the next generation of purpose driven AI talent.
• Claude Corps pairs early career fellows with nonprofits such as museums, social services, refugee support, and conservation groups to build AI tools that expand their impact.
• Fellows are paid full time for 12 months, receive training from CodePath, mentorship from Anthropic, and join a national cohort while hosts get a dedicated fellow, Claude API credits, and an implementation grant.
• Applicants must be over 18 with under two years of full time work experience and are selected based on AI experience, communication skills, and motivation to work on societal challenges.
• Anthropic is committing an initial 150 million dollars, with CodePath acting as employer of record and Social Finance leading measurement, evaluation, and long term funding design.
• The first 100 fellows will start in October 2026, with additional cohorts in January and August 2027, and the program’s curriculum and playbooks will be open sourced so others can replicate the model.
Claude Corps aims to ensure AI’s benefits are widely shared by channeling frontier tools into nonprofits and investing directly in early career workers. With applications for fellows due July 17 and host applications rolling, the program marks a significant bet on AI as a force multiplier for civic and social impact.
Hermes adds support for WhatsApp via Meta’s official Business Cloud API, a production‑grade option for customer facing bots.🗨️
No Node bridge, no QR codes, no personal account ban risk. Requirements: a Meta Business account, a dedicated business phone number, and a public HTTPS webhook (cloudflared, ngrok, or a reverse proxy).
Note Meta’s 24 hour customer service window replies after that require pre approved templates.
The setup wizard validates credentials and catches the common Phone Number ID mistake. For production use, create a System User permanent token to avoid daily token rotation. Configure Meta’s dev recipient whitelist and Hermes’ allowlist for safety. Install ffmpeg if native voice note TTS is desired.
Alternative: the Baileys bridge suits personal projects and quick demos but is unofficial and carries ban risk. For stable, compliant WhatsApp bots, the Cloud API is the recommended path.
Google is bringing one of the most requested features to the web version of Google Earth.✈️🧑✈️
Flight Simulator is now available globally for all web users, letting anyone take to the skies directly from their browser.
Google says it has also added several powerful desktop grade tools to the web version, including elevation profiles and new file import types, continuing its push to make professional mapping tools more accessible online.
But the headline feature is clearly Flight Simulator, a long awaited addition that blends exploration with pure fun.
Anthropic announced today that the US government issued an export control directive requiring suspension of access to Fable 5 and Mythos 5 for all foreign nationals, including foreign Anthropic employees.🚫⛔
To comply, Anthropic is disabling those models for all customers while access to other Anthropic models remains unchanged.
The government cited a potential jailbreak method that could bypass Fable safeguards.
Anthropic reviewed the demonstration and found a small number of previously known, minor vulnerabilities that other publicly available models can also surface.
Fable was developed with a defense in depth strategy and tested extensively with governments and third parties to reduce misuse risks.
Anthropic implemented 30 day customer data retention to support monitoring and rapid mitigation of jailbreak attempts.
The company says the government shared only verbal evidence of a narrow, non universal issue and that the capability demonstrated is widely available in other models used by defenders.
Anthropic is complying with the directive but disputes that a narrow finding should trigger a broad recall of a commercial model and warns that such a standard could halt future frontier deployments.
The company calls for a transparent, fair, technically grounded statutory process for blocking unsafe AI deployments.
Anthropic apologizes for the disruption and is working to restore access as soon as possible.
Expect service interruption for affected models while Anthropic and regulators engage.
Starting today you can bank Codex rate limit resets and use them when it suits you.
Go, Plus, Pro, and Business users get one free reset.
For the next two weeks, Plus and Pro can invite up to 3 friends. When a friend sends their first Codex message, you both earn another saved reset.
Everyone’s hyped about AI voices, but almost nobody is thinking about what happens when those voices get a consistent face.
What I like about this ElevenLabs Avatars update is that it quietly turns “talking head video” into a reusable content system for real teams:
Turn a script and avatar into a talking head video in one place
Upload photos once and keep a consistent on brand identity across angles and styles
Spin out assets for ads, social clips, and product videos without booking a shoot
Scale content across languages, scripts, and hooks inside the same workflow
Under the hood it’s still the same idea: take what used to be manual, one off production and wrap it in a repeatable flow you can run every day. You keep your brand, voice, and message; the avatar just becomes the face that shows up everywhere.
Coding isn’t “the future of work” anymore, how you think with AI is.🧠📈
We’re watching two things happen at once:
AI is quietly taking over a huge chunk of coding and even parts of software engineering
The real leverage is shifting to human centered skills: product sense, critical thinking, judgment, and knowing what actually matters to people
That’s exciting, but also dangerous. If you just offload everything to AI, you won’t just “move faster,” you’ll slowly deskill yourself. Students already let models write essays, devs let models write entire functions, and you feel productive while your own muscles quietly atrophy.
So if you’re 20–30 and trying to “future proof” yourself, maybe don’t obsess over learning to crank out boilerplate code. Learn to:
frame problems clearly
set context and prompt with intent
sanity check AI output with your own brain
navigate the messy, physical, human world AI can’t fully see
AI will happily do 95% of the work. Your edge is whether that remaining 5% is shallow or the part that actually requires judgment.
Watching the World Cup is only half the fun. The other half is building for it.
Prediction pools, ticketing, collectibles, payments → all shippable in one Saturday.
June 20 with @SuperteamMY and @KrackedDevs at @AWSstartups KL.
30 spots only 👀
The Anything CLI (@anything) is one of those tools that makes AI app building feel like a normal part of your dev workflow instead of a special event in the browser.
You install a single global binary (npm install -g @anythingai/cli), log in from the terminal or drop an API key into your env, and you’re ready to drive the same Anything agent you use in the UI directly from your shell.
From there, you can spin up and iterate on full apps with a couple of commands: create a project from a prompt, send follow up instructions, publish, roll back, and even pull the generated code down to your local machine.
Because every command can speak structured JSON and has flags for quiet output, non interactive mode, and clear exit codes, the CLI is built for scripting and AI coding agents as much as for humans, which means you can chain calls, wire it into CI, or let another agent orchestrate it.
Microsoft just sent a pretty loud signal about how seriously enterprises are about to take AI data governance.
They’ve reportedly limited internal use of Anthropic’s new Claude Fable 5 while their legal teams review Anthropic’s updated data retention rules, which keep prompts and outputs for 30 days by default, and up to two years when flagged for trust & safety review. That’s a big shift for a model Anthropic is positioning as a “safe for general use” public version of its powerful Mythos family, especially given its strengths in software engineering and analytics.
The concern inside Microsoft is straightforward: if employees use Claude Fable 5 on work systems, customer data and confidential information could end up stored on Anthropic’s side under those policies. Until their lawyers are satisfied, Fable 5 is effectively in a holding pattern for internal use.
👇
https://t.co/1HA0BEZSrt
Hermes can now call you when it matters.
Here's how it works:
You wire up Hermes with ElevenLabs’ ElevenAgents and Twilio (or another supported provider), then install the ElevenLabs skill so your agent knows how to talk to the voice APIs.
Instead of you calling your Hermes Agent, Hermes can now initiate the call using ElevenLabs’ outbound call API, with your agent ID, phone number ID, and your phone number in E.164 format.
From there, you can ask Hermes to generate a small watcher script that checks your calendar for meetings starting in a few minutes, builds a short prep brief, hits the ElevenLabs outbound call endpoint, and runs every minute via Hermes cron so it never double calls the same event.
The result is an autonomous Hermes Agent that phones you with context: “your meeting is in five minutes, here is the agenda, here is the decision you want, want to talk it through?”
The bottleneck never gets paid...
The layer above it does.
Proof:
ASML makes the only machine that can print an advanced AI chip.
5 years ~
*ASML +120%
*Nvidia +1,300%
Unavoidable ≠ Valuable.
So we mapped where value actually pools.
6 layers, bottom to top:
INFRA ~ datacenters + power. the physical floor.
CHIPS ~ the silicon. compute gets printed here.
DATA ~ the fuel. proprietary beats scraped.
MODELS ~ the intelligence itself.
EXECUTION ~ runtime + agents. where models touch reality.
APPS ~ the surface humans actually use.
The rule: own more layers, get harder to kill.
@amazon + google run all six
@OpenAI + anthropic hold three
@Lovable + figma ship on one ~ fast, brilliant, exposed
Every era of computing was won by whoever read the stack right.
map below 🟩