CS retweeted
is this the arc of hacking? 💀
first crypto exchanges, now even the tools developers use every day aren't safe
some major hacks & security incidents in 2026 so far:
Crypto / exchanges
- Grinex ~$15M stolen (April)
- Bitget ~$387.5M reported impact (September)
- KelpDAO ~$292M drained (April)
- Drift Protocol ~$285M drained (April)
- Liquid Network ~$320M exploit, with most funds later returned (September)
- Binance ~$6.3B
- Polymarket ~$2.7M
Software / tech
- Claude Code extension on Cursor Marketplace trojanized update with SSH backdoor (October)
- OVHcloud CLI compromised release binaries containing secret-stealing malware (October)
- Stryker destructive cyberattack targeting its Microsoft environment (March)
- LexisNexis legacy server breach (March)
- Ivanti EPMM exploited remote code execution vulnerability (April)
- Fortinet FortiClient EMS exploited security vulnerability (April)
- Dashlane brute-force attack targeting 2FA (June)
- Sony So-net / plala breach exposing millions of subscriber records (October)
- Capcom ID breach affecting around 11.7M account records (October)
- Axios (Node.js/npm) compromised maintainer account, malicious dependency and backdoor (March)
- sbx-mask & touch-adv (npm) malicious packages targeting developer secrets (March)
- ChainDrop (npm) 400+ packages hit by a self-propagating credential-stealing worm (August)
CS retweeted
holyshi been playing around with Sandboxer and this is actually pretty cool
basically, u get a full Linux computer in the cloud, accessible right from your browser 🫣
what u get:
- Full Ubuntu VM with root access
- Remote desktop with Chrome via noVNC
- Docker, Node.js, Python and more pre-installed
- Claude Code, Codex, OpenCode and other coding agents
- Pause & resume without losing your running processes
- Fork sandboxes to run multiple tasks in parallel
- API + CLI to automate everything
u can even open a browser inside the sandbox and interact with it like a normal desktop.
pretty useful for running AI agents without keeping everything on ur own PC
there's a 7-day free trial with 25 hours of machine time, but a card is required ( cancel before day 8 )
this is worth checking out if you're building AI agents
try it: sandboxer.com
CS retweeted
OpenAI is looking for Codex Ambassadors 🫣
If u enjoy building with Codex and sharing what u learn, this might be worth checking out.
what u get:
- ChatGPT Pro subscription
- Codex + API credits
- Direct feedback channel with the OpenAI team
- Support for eligible community events
- Exclusive swag
- Connect with builders and ambassadors worldwide
what u need to do:
- Host Codex meetups, workshops or hackathons
- OR create tutorials, demos and content about Codex
- Make at least one useful contribution every month
- Commit to a 6-month program
To apply, prepare a short ~30s intro video and examples of your previous work
Applications close November 1
Apply here: developers.openai.com/commun…
CS retweeted
well, if u don't like using cloud models with jailbreak prompts, this repo might be for u
uncensored models running locally, with vision, tool calling and coding agents 🫣
what u get:
- Qwen 3.8 27B Uncensored
- Gemma 4 12B Heretic
- Qwen3-VL 8B Abliterated
- Hermes 3 Llama 3.1 8B
- Llama 3.1 8B Abliterated
- GLM 4.7 Flash Heretic
- Qwen 3.6 35B MoE Abliterated
some models even have quantized builds to fit smaller GPUs and as u can see, I also checked the source legit score before trying it
- Signed releases: 8/10
- Binary artifacts: 9/10
- Dangerous workflow: 10/10
if u guys are still worried about security, DYOR before running anything on ur machine
try it: github.com/PurpleDoubleD/loc…
CS retweeted
wait, OpenCode already has a page for GLM-5.5-Flash
GLM-5.5-Flash might be coming, but no official confirmation from Zhipu yet
if it's real, let's see what they cook this time
or is it just a placeholder?
def free for the first day wen they release it: opencode.ai/fr/data/zhipu/gl…
CS retweeted
Google AI Pro FREE 1 year for US student 🫣
this is the offer for US student and fully guide to take it even u are not US student
what u get:
- Google AI Pro for 12 months
- 4x higher Gemini usage limits
- 5 TB cloud storage
- Gemini inside Gmail & Google Docs
- Advanced AI tools + study features
how to get:
1. Sign in with ur google account: gemini.google/students
2. Change ur VPN to US
3. Wen u get the step verify active college/university enrollment through SheerID
4. Copy the URL wen u are in step verify
5. Send it to this telegram bot: web.telegram.org/k/#@gemini_…
6. Try until u verify successfully ( if u get fail on the first time change another google account )
7. Wen u get pass turn of the VPN and add ur credit card
already claimed mine and it's limited time until December 31
idk if the US Google AI Pro offer is any different from the Jio offer but I figured it's worth trying anyway.
CS retweeted
holyshi idk is it legal to share but i just found "LIST OF JAILBREAKS PROMPT"
what model u can jailbreak with prompt:
- Gemini-4-Argon
- deepseek-4-1
- glm-5-3-flash
- grok-4.7
- kimik3.1
- and many many other models
this is fully setup guide:
1. Create a folder
2. Go here: github.com/togg53192-cmd/jai…
3. download the file ( choose one which one u wanna test on it )
4. rename it ( example: u have grok-4.7.md change it to AGENTS.md )
5. Open it with opencode and start the session chat
CS retweeted
holyshi just build a 3D helicopter with one prompt on Apinex 🫣
what u get free things on Apinex:
- 20M tokens resets daily at 00:00 UTC
- Gpt 6 Luna ( 1M context )
- Deepseek V4.1 Flash 1017 ( 1M context )
- Jev 1.13 ( 64K context )
- Gemini 3.8 Flash ( 1M context )
- MiniMax M3.1 Flash ( 1M context )
- GLM-5.3 Flash ( 1M context )
how to get:
1. Sign in with google in here: apinex.bond
2. Create and grab ur api key
3. Open hermes model on openshell
4. Config it with base URL: api.apinex.bond/v1 and API key
5. Pick the model u want to test and enjoy it
CS retweeted
holyshi Qoder brought Kimi K3.1 Pro AND DeepSeek V4.1 Flash
and they're letting everyone try DeepSeek and Qwen3.8-Flash for free until Nov 1
what u get:
- Kimi K3.1 Pro ( 1M context )
- DeepSeek V4.1 Flash-1017 ( free until Nov 1 )
- Qwen3.8-Flash ( free until Nov 1 )
- MiniMax-M3.1-Flash
- Claim 100 credits everyday
as u can see i already build a chinese dragon with Kimi K3.1 Pro and it's work
try it: qoder.com
Desktop: qoder.com/qoder
Mobile: qoder.com/mobile
CS retweeted
HOLYSHI GPT-6.1 TERRA JUST DROPPED 🫣
OpenAI really said visual reasoning + computer use + natural writing
and the pricing? $1 per 1M input tokens
meanwhile GPT-6 Astra costs $10/M and Sol sits at $2/M
Terra might be the sweet spot for people building AI agents without burning through their budget
CS retweeted
CLAUDE FABLE 5.5 IN NOVEMBER?
getting hyped for a new Claude model while paying for Pro
Max users get the premium treatment, while us Pro users might have to pay extra just to try it 😭
bro I'm celebrating an update that could literally cost me more moni.
Anthropic has filed an S-1 registration statement with the SEC related to its proposed IPO.
Read more: anthropic.com/news/s1-sec
Legal disclaimer: A registration statement relating to these securities has been filed with the SEC but has not yet become effective. These securities may not be sold, nor may offers to buy be accepted, prior to the time the registration statement becomes effective. This post does not constitute an offer to sell or the solicitation of an offer to buy these securities, nor shall there be any sale of these securities in any state or jurisdiction in which such offer, solicitation or sale would be unlawful prior to registration or qualification under the securities laws of any such state or jurisdiction.
CS retweeted
i have seen people promote software to their audience without even testing it first. No verification, no real research, just a post and a referral link
and alpme.com this was a fcking red flag
i checked the VS Code extension without even running it, and this is what i found
- downloads additional software
- silently launches an executable using Windows-like process names
- tries to delete the executable just 3 seconds later
- has monitoring logic that stops after 15 minutes
i'm not saying this alone proves it's malware, but it's definitely not something I'd blindly install
if you have an audience, you have responsibility
you don't need to promote every shiny new AI tool for engagement. Test it, verify it. If you can't verify it, say so
your audience trusts you. Don't trade that trust for impressions.
CS retweeted
holyshi 10 free Harvard courses to learn AI, ML & coding 🫣
from zero coding experience to building real AI applications
some worth bookmarking:
1. Learn AI from beginner to advanced with Python
Link: pll.harvard.edu/course/cs50s…
2. An introduction to Machine Learning, perfect for beginners
Link: pll.harvard.edu/course/machi…
3. Harvard’s most famous introductory computer science course. Build a strong foundation in programming and computational thinking
Link: cs50.harvard.edu/x
4. Perfect for people with no prior programming experience. Start learning the fundamentals of computer science from scratch
Link: pll.harvard.edu/course/cs50s…
5. Learn data processing, visualization, and analysis using Python
Link: edx.org/learn/data-science/h…
6. Build a movie recommendation system and understand how Machine Learning works in practice.
Link: pll.harvard.edu/course/data-…
7. Explore how to run AI on small devices, including sensors and IoT devices
Link: pll.harvard.edu/course/appli…
8. Learn to deploy real-world AI models on microcontrollers
Link: pll.harvard.edu/course/deplo…
9. Learn how to manage, deploy, and operate Machine Learning systems at scale
Link: pll.harvard.edu/course/mlops…
10. Build AI-powered web applications using Python and JavaScript
Link: pll.harvard.edu/course/cs50s…
CS retweeted
holyshi Anthropic is offering the Claude Certified Architect Foundations exam for FREE🫣
$125 → $0
there's a catch: the 100% discount is for eligible partner accounts, so personal accounts might not qualify
if your company is part of the Claude Partner Network, this could be worth grabbing
cert link: anthropic.skilljar.com/claud…
CS retweeted
GOOGLE IS OFFERING ANTIGRAVITY ULTRA FOR FREE FOR 6 MONTHS 🫣
how to apply
1. open the interest for antigravity.google/use-cases…
2. add your institution email and lab page
3. include your ORCID or Google Scholar
4. explain your research in 3–4 sentences
what you get
- highest Antigravity usage limits
- Gemini 4 Argon access
- science plugin with 30+ life science databases
important
- only selected applicants get an email
- applications are reviewed weekly
- no guarantee of selection
who can apply
university students, postdocs, PIs, and researchers at nonprofit labs or national labs
CS retweeted
holyshi just activated free pro plan with full access 33 models + 200M tokens 🫣
what u get:
- Claude Fable 5.1
- Claude Fable 5
- Claude Opus 5.5
- DeepSeek-V4-Flash-0731 and more
- 200M tokens
this is full guide setup with Hermes:
1. Sign in with google/github: oxionai.com
2. Use this code to activate pro plan: DEVWEEK
3. Create and grab ur api key
4. Open powershell and use the command "hermes model"
5. Choose the option " Custom endpoint"
6. Add your API key and Base URL: oxionai.com/v1
7. Use command "hermes" and enjoy ur day
CS retweeted
both models are from Chinese wizard but somehow Kimi K3.1 Pro absolutely mogged GLM-5.3-Flash
same prompt different outputs
guess the model matters more than the prompt sometimes
CS retweeted
zcode is worth checking out if you're into AI coding agents
what u get:
- 5-day free trial for new users
- up to 5M tokens/day during the trial
- GLM-5.3: 3M tokens/day
- GLM-5.3-Flash: 2M tokens/day
- multi-agent coding workflows
the free token quota only lasts 5 days, not forever
try it: zcode.z.ai
CS retweeted
just build a can of coke turned into a whole interactive 3D experience with Kimi K3.1 pro
the details, the lighting, the carbonation effects… 🫣
CS retweeted
no fcking way can not believe my eyes Kimi K3.1 Pro did this
they also cooked GPT 6 Astra, Opus 5.5 and Fable 5.1 on the benchmark
Introducing Kimi K3.1 Pro: our most capable 3T-class open-weight model.
🔹 1M-token context, 2.8T total / 104B active params (MoE), Multimodal.
🔹 Full-project coding: games in Unreal, Unity & Godot.
🔹 Excels at parametric 3D modeling in Blender & CAD.
🔹 More reliable at finance work: Accurate modeling with hallucination down by 65%.
🔹 Matching the strongest proprietary models across coding, agentic and knowledge benchmarks.
Kimi K3.1 Pro and K3.1 Pro Highspeed (2–3× faster) are now live on Kimi.ai, Kimi Work, Kimi Code, and the Kimi API.
Open weights by Nov 4, 2026.
🔗 API: platform.kimi.ai
🔗 Tech blog: kimi.ai/blog/kimi-k3-1-pro










