founder: https://xitter.gq/t.co/pfs8VtZMp6, blog: https://xitter.gq/t.co/GPZsnWkPlI, linkedin: https://xitter.gq/t.co/egLTe3tmyq
Palo Alto
Joined August 2008
- Tweets905
- Following74
- Followers900
- Likes195
Nishant Soni retweeted
Security shouldn't require a sysadmin.
This week, two critical PHP vulnerabilities were disclosed (PHP2Shell, CVE-2026-109618 and CVE-2026-109619). They allow attackers to take over a web server without logging in, and exploitation started within hours of disclosure.
Here's how we responded at NonBioS:
✅ Patched PHP across our customers' VMs
✅ Updated our base images so every new VM starts out safe
But the part we're most excited about is this: our users don't need to know any commands to protect themselves. They just tell nonbios what they want: "Check my VM for PHP2Shell and patch it if needed."
The agent checks the installed PHP version, applies the fix, restarts the web server and confirms the result.
And this won't be the last vulnerability. New ones are expected to come out rapidly across PHP, Apache, WordPress and countless plugins. The time between disclosure and active attacks is now measured in hours, not weeks.
So our advice to every builder is simple: when you hear about a new vulnerability, ask nonbios: "Research [vulnerability or CVE] and check if my VM is affected. If it is, fix it and tell me what you changed."
Staying secure should be as easy as asking. 🛡️
Nishant Soni retweeted
I've been thinking about Meta's impending Muse India launch.
I'm sitting on an invite code waiting to try it out, but I can't help but think that when Meta launches Muse in India, this may lead to a few things:
1. Calls for regulations of agents. Muse, my guess is, will be the trigger. Leave alone that ChatGPT, Claude, OpenClaw and Hermes (I am now running 6 hermes agents btw) have been around for a while, but Meta brings crazy scale with it. Razorpay has launched Ray as a whatsapp agent for merchants.
Muse will be the trigger because it has distribution, and Meta has cracked ease of use it seems, as has Instinct. But more than anything, it's because it is Meta, and of late, different arms of the government have been going at Meta (not without reason in many cases, but the intensity is hard to understand, and it reminds me of how they went after twitter in 2021 in order to justify weakening of safe harbor. They're scaring every platform into compliance, so they can exert more control over speech).
So, the moment Muse launches, someone will raise issues about agents with MEITY, and MEITY will be under pressure to say it's doing something to regulate agents.
Rajeev Chandrasekhar will write an oped on Muse and the need for regulating agents. Same guy who asked, in a MEITY PPT no less, why we need safe harbor, knowing fully well that no platform or ISP can survive the loss of safe harbor. Same guy who issued and defended an advisory that said AI models need to be registered to operate in India.
2. Concerns are going to be raised about Meta's data collection and the usage of this data for Meta's advertising business. Note that ChatGPT and Gemini are cheap/free in India, train on Indian data, and they have ad businesses. Btw, even some courts are using ChatGPT for judgments it seems. I see AI language in govt press releases. How are summaries of cabinet meetings created? Our govt needs to get its shit together wrt national security and AI usage by govt officials, because OpenAI has a deal with the Dept of War. Instinct's data collection issues have been raised recently, but it's not Meta in India.
3. Liability of agents will become a discussion. My view is, and I'm doing a research paper on agents, that agents have 3 body problem: model, harness (self-learning) and memory each bring variance to behaviour, and they are interdependent. You cannot have deterministic outcomes. Agents will make mistakes, as much as harnesses keep evolving to contain mistakes. How will we deal with this? I'm not sure.
4. Concerns may be raised about how easily these tools can game the system: for example, book tickets on IRCTC, thus putting humans at a disadvantage. This is especially since Meta seems to have cracked speed, and browsing. As has Instinct btw. Muse it currently tops assistantbenchmark.com with Instinct not very far, and I've been amazed by Instinct and how effortlessly it works.
5. Autonomous activity is going to freak people out and journalists will make headlines out of autonomous behaviour. I was pleasantly surprised when Instinct read my email, realised that I have got a payment notification from OpenCode for my monthly subscription, there's no receipt, and I had complained to opencode about payment failures (thank you RBI) in the past. All on its own. It's freaky and amazing at the same time.
On another note, I'm experimenting with setting up an autonomous OS for Reasoned (calling it ROS), where a majority of the work is done by agents autonomously, and I largely monitor. Starting with approvals, and after a few weeks of training Ed and Ev (two of my agents), I'm going to experiment with full autonomy. I've had Saki (my Chief of Staff agent) code autonomously for my in the past, and now each of my seven agents autonomously learn something new each night.
6. Some people will try and get agents to do things just so that they can raise an issue about what the agent did or said. We've seen this happen with chat outputs, going back to Bard in 2023.
Broader problem is:
It's really early for agents, and most people don't know how they work, where they fail, how they're built, how wonderful they can be, and how they change our lives for the better, how something breaks (every day) and how frustrating that can be, and while we've had a lovely year so far with agents empowering us and making things better for some of us, we're possibly going to see restrictions coming our way. Most people wanting agents to be regulated are operating out of ignorance and fear because they don't know how they work.
This is not about Meta, Google, ChatGPT, Anthropic, or even OpenClaw, Hermes, Instinct or the over 100 personal bots out there. Muse shouldn't be the trigger for panic. I want to keep using my hermes agents, and maybe set a few others up (11 at last count btw... 4 I'm not using)
My hope is that regulations are not drafted in ignorance, for political benefit, or speech control, or as a knee jerk reaction to people freaking out, but as with everything MEITY and Govt of India, and their feudal outlook on society, that's likely to be the case. My take here is: agents should have safe harbour protections, and due diligence should be on a best effort basis, else no agent will be able to function in India.
Someone repeatedly asked at a closed door discussion a couple of days ago: How do you respond when an agent goes rogue? Nobody knows. Regulation won't stop them.
Nishant Soni retweeted
We've noticed a user pattern we didn't design for, and it's become our most active segment.
They're not building a SaaS. They're building four.
Technically capable, distribution-savvy, and clear-eyed about the fact that most products don't work - so they ship several and let traction decide. We call them Portfolio Hackers.
Two features shipped today for them:
1. Multiple workspaces. One workspace per project. All chats for a project stay inside it. No more one giant list of thirty conversations across four unrelated products. Each workspace is tied to a computer - today the default VM, soon a machine of its own, and after that your own VM from Linode or DigitalOcean via BYOC (Bring Your Own Computer).
2. Start/stop your computer. Directly requested by our heaviest users, who keep some machines for development and others for hosting. Turn off what you're not using. Free-plan users can also boot and warm up the machine before handing work to nonbios.
A single $9 NonBioS plan is a 4GB/2vCPU VM with a public IP and MySQL preinstalled - enough to launch multiple SaaS apps and keep them running while you figure out the market. Now the interface finally matches how portfolio hackers actually work.
Live now for everyone, free plans included.
Nishant Soni retweeted
Pricing and Infrastructure Update: Longer-Lived Free Computers, More Powerful Paid Plans, and Long Horizon Autonomy
Running general-purpose Computers for autonomous AI agents comes with a less glamorous operational reality: people will try to abuse them. We have seen a remarkable range of it: Botnets. Phishing infrastructure. Cryptocurrency mining. Automated account abuse.
We never wanted the answer to be simply making the Free tier less useful for everyone else. Instead, we’ve been working on better ways to identify and contain abusive workloads while giving real builders more freedom.
Previously, every NonBioS user received essentially the same Computer: 4 GB RAM, 2 vCPU. Free Computers were deleted after 3 days of inactivity. Paid Computers stayed around indefinitely for as long as the subscription remained active. That worked, but three days turned out to be too short for many people.
So we’ve changed the Free tier.
Free users still get a Temporary Computer, but your files now remain available for up to 7 days of inactivity. The Computer may stop after a period of inactivity, but you can restart it simply by initiating another chat. If you return within seven days, your project is still there.
Paid plans now get a more Powerful Permanent Computer
NonBioS is not just an AI interface sitting on top of somebody else’s hosted project environment. The idea behind the product has always been simple:
An AI Software Dev with its own Computer.
That Computer becomes more capable as you move into higher plans. Higher tiers can provide additional RAM, CPU, disk capacity, and longer autonomy horizons for projects that need them. You can run multiple applications, integrations, databases, APIs, workers, and domains on the same machine, subject to the actual resources available on that Computer.
Legitimate users get more time, not less: projects now stick around for seven days of inactivity instead of three. At the same time, infrastructure that isn’t being actively used doesn’t need to remain online indefinitely.
The other major change came from one of the most common requests we’ve received: let NonBioS work for longer on its own.
NonBioS is designed to work autonomously, but autonomy needs sensible boundaries. Today, while working on a task, NonBioS can complete a large predefined number of autonomous steps before it automatically pauses and waits for you to review what it has done. You can then inspect the work, give feedback, or simply tell it to continue.
But something interesting happens over time: People who have used NonBioS for a while tend to want fewer interruptions.
That is what Long Horizon Autonomy is for. Higher plans now increase the number of autonomous steps NonBioS can take before the automatic review pause.
Full details at nonbios / pricing
Nishant Soni retweeted
We just bumped credits again. A few weeks ago we nearly doubled credits across all plans. Today we're doing it again.
Here's where things stand now:
Starter: 80 → 150
Vibecoder: 300 → 500
Developer: 1,200 → 1,750
Architect: 2,500 → 3,600
This is part of our ongoing infrastructure push - we said we'd keep passing gains directly to users, and we meant it.
More headroom, more often. You should always be able to build !
Deets at nonbios.ai/pricing
Nishant Soni retweeted
Yasin Malik’s men didn’t stop at raping the 27-year-old Kashmiri Hindu nurse Sarla Bhatt. They then murdered her, engraved the acronym ‘JKLF’ on her body, and mutilated it out of sheer contempt for her Hindu identity. After that, they issued multiple warnings to her family against cremating her remains according to Hindu rituals. The family still went ahead with the cremation, defying the threats. When they went to collect her cremated remains (asthi) later, a mob of around 200 Islamists descended upon the crematorium and stomped over them.
They didn’t stop even at that. They then bombed her family home with a grenade.
Do you really think all 200 of those monsters who showed up at the crematorium were foreign infiltrators, Pakistani terrorists, or JKLF members? These are the questions that make many Kashmiri Muslims very uncomfortable. The moment you raise these issues, especially when they present a narrative portraying their community as victims, they label you an Islamophobe or resort to similar accusations.
I have seen all kinds of Islamists, but the tribe in Kashmir is the craftiest. Many of them pretend to be atheists in private, yet their public behaviour and political leanings tell a completely different story. Very hard to take their words at face value.
Nishant Soni retweeted
Pricing update: We’ve almost doubled the credits across all NonBioS plans enabled by our recent infrastructure updates and feedback from users.
Also, note that nonbios-1.143 is now selected by default. It’s now labeled as Premium Coding Agent - and consume credits faster.
To get more mileage from your credits, switch to the Standard Coding Agent - nonbios-1.137 whenever you start a new chat.
Full Details: nonbios.ai/pricing
Nishant Soni retweeted
We just shipped something we've wanted for a while: Daily Credits on NonBioS.
Until now, running out of plan credits before your billing cycle renewed meant hitting a wall. That never felt right, you shouldn't lose momentum on a project just because you're between cycles.
So now every plan, including Free, falls back to a daily credit allotment that resets every 24 hours. When your plan credits run out, you keep going at a steady pace until your cycle renews or you top up. No lockouts, no surprises.
It's a small change, but it reflects a new way of thinking at NonBioS: You should always be able to build.
Check it out -> nonbios.ai/pricing
Nishant Soni retweeted
found something cool for someone who's into tech, AI, productivity, building in public or vibe coding.
and what if AI could actually build the whole thing for you?
i recently checked out @nonbios and what caught my attention wasn't the AI coding part cos we've seen plenty of that already.
what stood out to me is that they don't stop at generating code. it can install dependencies, fix errors, manage the environment and even deploy the final product live on its own.
imagine having an app idea at 2am, instead of spending weeks learning frameworks, watching tutorials, setting up servers and debugging random errors, you simply describe what you want and let the AI handle the heavy lifting.
the fact that it runs on a real linux machine with a public IP makes it even more interesting because it's working in an actual environment, not just inside a chat window.
what i also like is their focus on long projects. most AI tools are great until the project becomes too large and they start forgetting what they're building. NonBioS claims to solve this with something called strategic forgetting allowing it to work autonomously for hours without losing track of the goal.
tbh, people have reportedly built marketplaces, gaming platforms, community portals, SaaS products, APIs and even meditation apps without writing code themselves.
we're getting closer to a world where having an idea matters more than knowing how to code.
not saying developers are going away anytime soon, but tools like NonBioS are making it much easier for creators, founders and builders to turn ideas into real products.
for $9/month, that's a pretty interesting to me
have you tried letting AI build something completely from scratch yet?
Nishant Soni retweeted
One of the biggest frustrations with AI coding tools is that they can build 80% of something...
...and leave you figuring out the remaining 20%.
That's usually where the real work begins.
@nonbios is taking a different approach.
Instead of acting like a coding assistant, it acts more like an engineer living inside your browser.
You describe what you want.
It:
• Writes the code
• Installs dependencies
• Fixes errors
• Deploys the project
• Keeps working autonomously on larger tasks
A few things I found interesting:
🔹 Build apps, SaaS products, APIs, and tools
🔹 Handles deployment for you
🔹 Runs on real infrastructure
🔹 Designed for long-running complex projects
🔹 Accessible even if you can't code
We're reaching a point where building software is becoming more about ideas than implementation.
Tools like NonBioS are pushing that shift even further.
#AI #NoCode #VibeCoding #BuildInPublic #Startups
🤝 Paid partnership
Nishant Soni retweeted
Built entirely on NonBioS: PlayersAreGamers.nl
A full real-time multiplayer gaming platform. Compete for coins, climb the global leaderboard, and win real prizes. Wallet system, live matchmaking, the works.
No dev team. No months of work. Just nonbios.
📺 Watch it in action: youtube.com/@playersaregamer…
Nishant Soni retweeted
great honest assessment of OpenClaw by @sonink. after running an OpenClaw agent for 90 days, there's nothing I disagree with. if you like tinkering and peaking around corners, OpenClaw is an incredibly fun but legitimate use cases today have an effective EV = 0.
open.substack.com/pub/nishan…
Nishant Soni retweeted
What if "easy to learn" is actually expensive?
Bubble(.io)'s gentle learning curve hides a trap: proprietary skills that expire the moment you leave.
Linux/NonBioS AI feels harder upfront, but every hour invested pays dividends forever.
The full story: nonbios.ai/best/bubble-alter…
Nishant Soni retweeted
Less than 24 hours since launch of the new models, and the feedback is already coming in.
Two new models: nonbios-1.135 and nonbios-1.141
We're releasing two experimental models today that represent significant Context Engineering Architectural breakthroughs in AI-powered software development.
But first, let me explain the foundation these models are built on.
Strategic Forgetting: Less Memory, More Focus
Most AI agents try to remember everything. We took the opposite approach - we invented Strategic Forgetting, an algorithm that continuously prunes an agent's memory to keep it sharp and focused.
Think about how you work. When you're deep in debugging a complex issue, you naturally filter out background noise - the conversation happening across the room, the email notification, the tangential code comments. You keep what matters for the task at hand and let everything else fade. That's exactly what Strategic Forgetting does.
Today's releases represent a fundamental reimagining of how context engineering works within Strategic Forgetting. We've built a completely new architecture that doesn't just prune better - it understands better.
nonbios-1.135: Speed Meets Intelligence
Nearly 2x faster than nonbios-1.13, but speed is only part of the story. We achieved this through two fundamental innovations. First, we parallelized key parts of our Strategic Forgetting algorithm, yielding a 25% speed boost.
But the real breakthrough came from our new context engineering architecture. This new architecture helps nonbios-1.135 "converge" on correct solutions dramatically faster. It remembers key details twice as well as 1.13, which means it not only works faster, it solves problems where 1.13 would simply give up. The cumulative effect is roughly 2x performance on software engineering tasks.
And since we charge by the minute, faster execution means more bang for your buck.
nonbios-1.141: The Bug Hunter
Built on Claude Opus 4.6 and incorporating the same architectural advances as 1.135, nonbios-1.141 is something different entirely. In our testing, it solved complex bugs that every other model in our lineup failed on. Not some bugs. Not most bugs. Every single challenging case we threw at it. Even we were shocked by the results.
Here's a real example: We had a complex React bug that nonbios-1.13 struggled with for almost 10 hours, proposing multiple wrong solutions along the way. nonbios-1.141 isolated the exact issue in 5 minutes and one-shot the fix.
But 1.141 isn't perfect. On another React bug, it initially proposed the wrong solution but when we pointed out the inadequacy, it quickly converged to the correct fix.
The combination of our new context engineering architecture and Claude Opus 4.6's capabilities creates debugging performance we haven't seen before.
There's a tradeoff - 1.141 is dramatically slower than other models. But when you're hunting a critical bug that's been burning hours or days, speed takes a back seat to actually solving the problem.
Nishant Soni retweeted
New Tutorial: Setting Up Git & GitHub for Your Web App
Ever lost code because you didn't have proper version control? We've all been there.
In Episode 2 of Vibecode Your SaaS, we show you how to set up Github using NonBioS.
What you'll learn:
1. Navigate your server file system with a web-based file browser
2. Understand Linux directory structure (/home vs /var/www)
3. Initialize Git repos without memorizing commands 4. Create GitHub Personal Access Tokens
5. Push your first commit to GitHub
Pro tip from the video:
Create separate AI chats for different tasks:
- One for development
- One for GitHub operations
- One for deployment
They all access the same VM, keeping your workflow organized and context clean.
We're building a complete Health Tracker SaaS from scratch. Episode 2 adds the professional foundation every production app needs: proper version control.
No more "final_v2_ACTUAL_final.php" 😅
youtube.com/watch?v=JXGAd3ms…
Nishant Soni retweeted
Proud to share how Prateek (linkedin.com/in/iamprateeksr…) built EnglifyAI - an AI-powered grammar & writing assessment tool for Indian learners - using NonBioS! 🚀
From ideation to deployment on Vercel, NonBioS handled the heavy lifting: code generation, testing, evals & more.
✨ What makes this special:
- Built for Bharat, by Bharat
- GenAI-powered Written Assessment Test reports
Complete end-to-end development with autonomous coding
🎥 Demo: drive.google.com/file/d/1bW3…
💻 GitHub: github.com/Prateek-DSB/engli…
