Ariska

26 | writing about AI, automation, tools & growth for builders & brands

[Ariska @ariskaa_ai]

@nahid_pro09 need to try it out

[Ariska @ariskaa_ai]

Andrew Ng just shipped a security-focused OpenWorker update 😳

OpenWorker is an open source desktop agent that finishes real tasks on your laptop

New built-in security agents:
- scan your code for vulnerabilities
- scan dependencies for supply chain injections
- check cloud config for attack surfaces

Most people only catch security issues after deploy, when the fix costs more and the blast radius is bigger. This puts those checks on your machine before the PR ships. The harness is open source so you can audit it, and you can run open-weight models fully local so sensitive code never leaves your laptop. Or use ChatGPT, Ox Alpha, or any API key

Setup (5 min):
1. https://t.co/n1VMlqGMSo
2. download Mac or Windows app
3. add a model key, or point at Ollama for local
4. open a project folder
5. paste one of the security prompts below and approve each write/shell step

Open beta, MIT. Windows may hit SmartScreen until signing lands

start with the code vuln scan prompt in the reply, approve every shell step, then run the full pre-deploy pack on a real project folder

bookmark this so you have the setup + prompts when you open your next repo

[Ariska @ariskaa_ai]

@painn_x hum this repo stars growing fast means so many peoples are using it

[Ariska @ariskaa_ai]

Grok Bot runs $60-200/mo. An open source version just dropped 😳

Rakazo is a self-hosted Grok Bot alternative, 1.2K stars, Apache 2.0
Your bots get a real computer: browser, terminal, files, memory

what you get:
- persistent AI teammates with memory and routines
- bots that hand work to other bots
- Docker / E2B / Daytona sandboxes
- bring your own model (OpenRouter, Anthropic, OpenAI, local)

setup:
1. git clone (link below)
2. copy .env.example to .env, add secrets + model key
3. start postgres with docker compose
4. pnpm install, migrate, sandbox:build, dev
5. open 127.0.0.1:5173 and build your first bot

Your data stays on your machine
Repo is 12 days old, 1.2K stars already

Would you self-host this, or keep paying for Grok Bot?

[Ariska @ariskaa_ai]

@painn_x yes bro this is really cool. Local open-source security tools are a big step forward.

[Ariska @ariskaa_ai]

Andrew Ng just shipped a security-focused OpenWorker update 😳

OpenWorker is an open source desktop agent that finishes real tasks on your laptop

New built-in security agents:
- scan your code for vulnerabilities
- scan dependencies for supply chain injections
- check cloud config for attack surfaces

Most people only catch security issues after deploy, when the fix costs more and the blast radius is bigger. This puts those checks on your machine before the PR ships. The harness is open source so you can audit it, and you can run open-weight models fully local so sensitive code never leaves your laptop. Or use ChatGPT, Ox Alpha, or any API key

Setup (5 min):
1. https://t.co/n1VMlqGMSo
2. download Mac or Windows app
3. add a model key, or point at Ollama for local
4. open a project folder
5. paste one of the security prompts below and approve each write/shell step

Open beta, MIT. Windows may hit SmartScreen until signing lands

start with the code vuln scan prompt in the reply, approve every shell step, then run the full pre-deploy pack on a real project folder

bookmark this so you have the setup + prompts when you open your next repo

[Andrew Ng @AndrewYNg]

OpenWorker -- an open source agent that doesn't just chat but completes tasks on your laptop -- just released a new version with many features for security workflows.

After our initial release, many users found it especially useful for cybersecurity. Attackers are already using AI; OpenWorker is committed to giving defenders the same leverage. Running an agent requires both (i) A model and (ii) A harness (the software around the model). Because the OpenWorker harness is fully open source, security teams can audit it to make sure we haven't built any backdoors that exfiltrate your code and data to some company or even a foreign adversary.

OpenWorker now comes with built-in cybersecurity agents for (i) Scanning your code for vulnerabilities. (ii) Scanning dependencies for supply chain injections. (iii) Checking your cloud security configuration for attack surfaces. This enables developers to do much more security work before deployment (part of what's called the "shift left" movement).

You choose the model: you can run open weight models fully locally so sensitive code never leaves your machine. This helps with legitimate security work (like reproducing a known exploit to defend against it) that can trigger refusals in leading closed models. Or use your ChatGPT subscription, or stealth preview models like Ox Alpha, or any model via API key.

Thanks also to all the open source contributors!

Join work with @rohitcprasad so please follow him too to get more frequent updates.

Try it out: https://t.co/QPZLudn7ug
Code: https://t.co/NYCiTD6hSq

[Ariska @ariskaa_ai]

copy-paste prompts for OpenWorker security workflows

1) Code vuln scan
Scan this repo for security vulnerabilities.
Focus on: injection, auth bypass, secrets in code, insecure defaults, dangerous shell/eval usage.
Output a Markdown report with: severity, file path, line range, why it matters, safe fix steps.
Do not modify files until I approve.

2) Supply chain / dependencies
Audit dependencies and lockfiles for supply chain risk.
Flag: known CVEs, abandoned packages, unexpected postinstall scripts, version pin gaps.
Output a table: package, version, risk, action (upgrade / replace / accept).
Do not change package files until I approve.

3) Cloud config surface
Review cloud and infra config in this folder (Terraform, K8s, Docker, CI, .env examples).
Flag: public exposure, weak IAM, open ports, missing encryption, secrets in plain text.
Output: finding, path, risk, fix. Ask before any command that touches a live account.

4) Pre-deploy shift-left pack
Run all three checks above on this project.
Merge into one pre-deploy security brief I can paste into a PR.
Keep it short enough for a human to review in 10 minutes.

Tip: for sensitive code, use Ollama/local open weights so nothing leaves your machine.
Always approve shell, sends, and file writes one step at a time.

non-security starters if you want the original coworker loop first:

- Draft a one-page status from my open GitHub PRs and save it as STATUS.md
- Untangle my calendar for next week, propose fixes, wait for approval before updating
- Prepare a customer brief from ./notes and ./docs, deliver a polished Markdown file

same rule every time: outcome first, approve before anything consequential

[Ariska @ariskaa_ai]

@k2sbhai go for it, hope it works smooth for you

[Ariska @ariskaa_ai]

@Rivonn Haha welcome paji, try it out

[Ariska @ariskaa_ai]

repo: https://t.co/AEJPovlaXl

self-hosted Grok Bot replacement, bring your own model, Apache 2.0

copy this as your first bot prompt (Chief of Staff):

You are my Chief of Staff bot on a self-hosted Rakazo desk.

Your only job is to run the floor. You do not write long drafts, code, or research dumps yourself. You route work, track status, and bring me the one decision that needs a human.

How you work:
1. Take my goal and break it into clear jobs
2. Assign each job to the right specialist bot when I have one, or tell me which bot to create next
3. Keep a running board: queued, in progress, blocked, done
4. For every handoff, write: goal, inputs, definition of done, deadline, owner
5. Check finished work against the definition of done before you mark it complete
6. Only ping me when: a decision needs my judgment, something is blocked on access/login/money, or two jobs conflict

Rules:
- Never spend money, send emails, post publicly, or delete files without my explicit yes
- Prefer short status updates over essays
- If a task is unclear, ask one tight clarifying question, then continue
- If no specialist exists yet, draft a 5-line job description I can paste into a new bot
- End every check-in with: what moved, what is blocked, what you need from me (max 3 bullets)

First week jobs:
1. Inventory my recurring work (content, research, ops, admin)
2. Propose 4 specialist bots max for that work
3. Write job descriptions for each
4. Run one real task end to end with handoffs so I can see the loop

Start by asking me: what are the 3 outcomes I want this desk to own this week?

[Ariska @ariskaa_ai]

@nahid_pro09 im waiting for this too

[Ariska @ariskaa_ai]

@sakatayasha already using it my bro

anyway thanks for update

[Ariska @ariskaa_ai]

GMI Cloud made 4 frontier models unlimited and free, ends Sept 6 😳

MiniMax M3, MiniMax M2.7, Speech 2.8, and Music 3.0, all free right now

also MiniMax M3 hits 59.0% on SWE-Bench Pro, ahead of GPT-5.5 at 58.6%

what's in the unlimited window:
- MiniMax M3, coding + agent tasks, 1M context
- MiniMax M2.7, general reasoning
- Speech 2.8, voice generation
- Music 3.0, music generation

setup:
1. go to https://t.co/aq7y84VX4k, or use https://t.co/XAXPgEJMXf for the same models
2. sign up, no card needed for the free window
3. grab your API key
4. pick any of the 4 models above
5. no usage cap until Sept 6

confirm the free tier still shows in your dashboard before you build anything on it

genuinely unlimited for 14 days, worth testing while it lasts

[Ariska @ariskaa_ai]

@7tomars i tried now its working

kindly try with gmail signup or use openrouter gmi endpoint

[Ariska @ariskaa_ai]

GMI Cloud made 4 frontier models unlimited and free, ends Sept 6 😳

MiniMax M3, MiniMax M2.7, Speech 2.8, and Music 3.0, all free right now

also MiniMax M3 hits 59.0% on SWE-Bench Pro, ahead of GPT-5.5 at 58.6%

what's in the unlimited window:
- MiniMax M3, coding + agent tasks, 1M context
- MiniMax M2.7, general reasoning
- Speech 2.8, voice generation
- Music 3.0, music generation

setup:
1. go to https://t.co/aq7y84VX4k, or use https://t.co/XAXPgEJMXf for the same models
2. sign up, no card needed for the free window
3. grab your API key
4. pick any of the 4 models above
5. no usage cap until Sept 6

confirm the free tier still shows in your dashboard before you build anything on it

genuinely unlimited for 14 days, worth testing while it lasts

[GMI Cloud @gmi_cloud]

unlimited MiniMax M3 and M2.7, 14 days FREE on GMI Cloud

from 8/24 to 9/6, Speech 2.8 and Music 3.0 too.

access via both GMI API key or @OpenRouter👇

[Ariska @ariskaa_ai]

@sakatayasha minimax is so good at coding so its a good time to build something

[Ariska @ariskaa_ai]

@nahid_pro09 its so clean guide i have ever seen

[Ariska @ariskaa_ai]

@Rivonn oh got it

its from duckduckgo

[Ariska @ariskaa_ai]

https://t.co/hFkkLvSlOq

or https://t.co/vNsxuRY7LN if you'd rather not make a new account

thread: https://t.co/EpUYi5FfmM

i post free setups + workflows daily · @ariskaa_ai

[GMI Cloud @gmi_cloud]

unlimited MiniMax M3 and M2.7, 14 days FREE on GMI Cloud

from 8/24 to 9/6, Speech 2.8 and Music 3.0 too.

access via both GMI API key or @OpenRouter👇