Forem: Soon Seah Toh

Six Million Singaporeans, the Productivity of Six Hundred Million: The Budget 2026 Bet

Soon Seah Toh — Mon, 20 Apr 2026 01:09:23 +0000

PM Wong just made the case for every Singapore company building AI. And he did it more directly than anyone expected.

In Budget 2026, he said: "Fear cannot be Singapore's response. If we allow uncertainty to paralyse us, we will fall behind."

That's not typical political language. That's a direct challenge.

And here's the part that should excite every Singaporean reading this.

If we get this right, the math changes everything for our country.

Singapore has always punched above its weight. Six million people, no natural resources, no domestic market, surrounded by giants. We've built one of the world's most successful economies on discipline, talent, and strategic positioning.

Now imagine layering AI super-productivity on top of that.

One knowledge worker doing the output of ten. The most senior ones doing the output of a hundred. And if we rally most of the population to that level, what does that look like nationally?

It looks like a country with the workforce of 6 million doing the output of 60 million. Or 600 million. Suddenly the size of our population stops being a constraint. Suddenly we are not the small player in the room. We are the most leveraged one.

This is the bet Budget 2026 just placed.

A new National AI Council, chaired by the PM himself
National AI Missions in 4 priority sectors: manufacturing, connectivity, finance, healthcare
An AI Park at One-North to catalyse ideas and collaborations
6 months of free premium AI tool access for Singaporeans completing training
Enhanced Enterprise Innovation Scheme covering AI expenditures up to $50,000/year
SkillsFuture redesigned with clear AI learning pathways

Singapore can't build frontier LLMs at OpenAI's scale. But we can deploy AI better than anyone else. Apply it harder. Ship it faster. Train our workforce smarter. That is the strategic advantage. That is the multiplier.

And the SG AI deployment story isn't a future ambition. It's already being written. Right now. By SG-incorporated, SG-headquartered, SG-built companies that have been quietly shipping production AI for years.

NetGain Systems is one of them.

We're a 23-year-old Singapore company that pivoted hard into AI for IT operations. Today we ship Astra AI — five specialist AI agents that autonomously investigate IT incidents, diagnose issues, and recommend fixes. Built in Singapore. Sold globally. Used by enterprises across Asia, Europe, and North America. We are exactly the kind of deployment company Budget 2026 is talking about.

A few days ago, I wrote that the real divide in the AI era will be builders vs bystanders. PM Wong just said the same thing at the national level.

To my fellow Singaporean founders, operators, engineers — the government just gave you tools, funding, and direction. If you are not moving on this now, you will be explaining yourself for a decade.

To the National AI Council, EDB, IMDA, JTC, and the agencies driving this — we are here. Singapore companies are already doing the work. Let's amplify it. Let's make sure the world sees what a small island nation can become when its people are armed with AI.

Six million Singaporeans, each with the productivity of ten or a hundred, is not a small country anymore. It is a powerhouse.

Fear cannot be our response. Ambition has to be.

Let's go build the future where Singapore stays strong and resilient — not in spite of our size, but because we used AI to redefine what size means.

Written by Soon Seah Toh, CTO & Founder of NetGain Systems.

The Secret Sauce Leaked. Turns Out There Was No Secret.

Soon Seah Toh — Sun, 19 Apr 2026 12:24:12 +0000

The best-kept secret in AI just leaked.

And it turns out there was no secret.

On March 31, Anthropic accidentally shipped the entire source code of Claude Code to the public npm registry. 513,000 lines. 1,900+ files. Unobfuscated. The crown jewels of the company that arguably builds the best AI agent on the planet, suddenly sitting in everyone's node_modules.

The internet did what the internet does. Mirrored it. Forked it. Dissected it. By the time Anthropic's DMCA takedowns went out, it was already everywhere.

Here's the part nobody saw coming.

When security researchers started publishing their analyses (Zscaler, IANS Research, half of dev.to), we all braced for some exotic architecture. Some proprietary trick. The secret sauce that explains why Claude Code leaves every other coding agent eating dust.

You know what they found?

Nothing fancy.

No magic. No proprietary algorithm. No hidden layer of sorcery. Just a ruthlessly well-engineered while loop with discipline around it that most teams skip.

I called this two months ago in a post saying the agentic loop is basically "Hello World" for AI agents. A lot of people pushed back. "It's way more sophisticated than that." "You're oversimplifying." "There must be something we're missing."

Turns out there wasn't.

Here's what the world's best coding agent actually looks like under the hood:

The loop is embarrassingly simple. Call the LLM. Parse for tool calls. Execute them. Append results. Loop until done. That's it. The part people keep trying to complicate is genuinely not that complicated.

The obsession is in the plumbing. Read-only tools run in parallel, up to 10 at a time. Write tools run one at a time, deliberately. This one decision alone makes Claude Code feel 10x faster than agents that don't bother. It's not a trick. It's an architectural choice that most teams skip because they're chasing features.

Safety is built as layers, not hoped for. Every tool call hits a permission check. Every input gets schema-validated. Write operations serialise by design. You don't tell customers your agent is safe. You make it structurally impossible for it to misbehave.

Context assembly is its own discipline. System prompt, project memory, git state, tool definitions — all carefully layered, not hand-waved. Auto-compaction kicks in before hitting the window limit. Boring. Mechanical. Ruthless.

Sub-agents are first-class citizens. Not some bolted-on afterthought. The coordinator module exists specifically to spawn specialist agents for focused tasks without polluting the main context. This is exactly what I wrote about back in February when describing how our 5-agent Dark NOC investigates incidents. This is how you scale agents. Most teams don't even attempt this.

Memory, skills, plugins — all separate modules. Clean boundaries. Each component does one thing. The unsexy engineering discipline that separates systems that last from frameworks that hit GitHub trending and then quietly disappear.

So what's the real lesson from the biggest accidental leak in AI history?

The secret sauce was never secret. It was just hard.

Anthropic isn't winning because of some proprietary trick they stumbled onto. They're winning because they have an army of engineers doing boring, disciplined, relentless work on the layers around a simple loop. The plumbing. The safety. The parallelism. The memory management. The sub-agent orchestration. The stuff nobody wants to talk about because it doesn't make for a viral demo.

Here's why I'm genuinely excited about this.

We've been building Astra AI at NetGain for over a year. Five specialist agents that autonomously investigate IT incidents, diagnose issues, recommend fixes. When I read through the public analyses of Claude Code's architecture, I had two reactions at the same time.

Relief. Because the patterns we arrived at independently are almost identical. The agentic loop. Tool parallelism. Layered permissions. Sub-agent orchestration. Memory separation. We got there on our own, because this is what actually works when you stop chasing hype and start shipping something customers will put their production workload on.

Motivation. Because we're not done. Claude Code has had hundreds of engineers polishing it for years. We're a smaller team moving fast. But the direction is right and the finish line looks closer than people realise.

If you're building AI agents right now, here's the honest advice:

Stop looking for the next framework. Stop chasing the next silver bullet. The fundamentals are on the table now. What separates great agents from toy demos isn't novelty. It's engineering discipline. Boring, focused, relentless execution on the layers around a simple loop.

The Claude Code leak was a disaster for Anthropic. For everyone else building in this space, it's the greatest gift the industry has ever received. The bar just got clearer. The mystery just got dispelled. The playing field just got levelled.

Nobody gets to hide behind "we have a secret architecture" anymore.

Go build.

Written by Soon Seah Toh, CTO & Founder of NetGain Systems.

AI Will Run Companies. Here's Why That Should Excite You, Not Scare You.

Soon Seah Toh — Fri, 17 Apr 2026 14:17:38 +0000

I'm going to say something that most people aren't ready to hear.

AI will run companies. Not help. Not assist. Run.

One founder. Zero full-time staff. Twenty AI agents handling support, dev ops, sales, marketing, finance, HR, and every system the business touches. Working 24/7. Never calling in sick. Never forgetting a process. Never dropping the ball at 2am.

"That's crazy."

No. What's crazy is that we already have every building block to make this happen — today.

What's Already Possible Right Now

Think about what's already possible right now:

Monday morning. Your AI operations team has already pulled data from GitHub, Bugzilla, Jenkins, and CRM. Weekly status report — compiled. Twelve internal support tickets — handled overnight. A draft email to leadership sitting in your inbox. You glance at it, change one word, hit approve. Done.

A customer emails at 3am with an urgent issue. The support agent reads it, searches your knowledge base, drafts a response, and holds it for your approval. By the time you wake up, all you do is tap "send."

A build breaks at midnight. The dev ops agent detects it, applies the standard operating procedure, rolls back, and leaves you a summary. You read it over coffee.

Now multiply that across every function in your company.

You Teach These Agents by Doing Your Job

Here's the part that should excite you: you teach these agents by doing your job.

Approve a reply as-is — the agent learns it got it right.
Edit something before sending — it remembers exactly what you changed and why.
Reject something — it never does it again.

You're not programming. You're mentoring.

After 50 approvals with zero edits, the system says: "This agent has a 100% track record on internal tickets. Want to let it run autonomously?" You say yes. That's one less thing. Then another. Then another.

Six months in, you're handling only the 20% that genuinely needs a human brain. Everything else just runs.

"But What About Control?"

You have more control than you've ever had.

Every external message needs your approval.
Every risky action needs your approval.
Full audit trail on every decision.
One-click pause on any agent.
Trust is earned over hundreds of interactions — never assumed.

Your AI team has more guardrails than most human teams do.

"But What About Jobs?"

Wrong question.

The right question is: what becomes possible when the cost of running a company drops by 80%?

How many ideas never got built because someone couldn't afford a team of ten?
How many founders gave up because operations crushed them before the product had a chance?

AI that runs companies doesn't kill opportunity. It democratises it. One person with a vision and twenty AI agents can now compete with companies that have fifty employees. That changes everything.

We're Just at the Start of This Curve

The LLMs today are the worst they'll ever be. Every month they get smarter, more reliable, more capable. If this is what's possible now, imagine where we'll be in two years. Five years.

The gap between "AI assisted" and "AI operated" is closing faster than anyone expected.

This is not science fiction. This is not a TED talk fantasy. This is buildable, today, with technology that already exists.

The only question is whether you'll be the one building it — or the one watching someone else do it.

Dare to dream this. Because it's coming whether we're ready or not.

Written by Soon Seah Toh, CTO & Founder of NetGain Systems.

Discover NetGain V15 — AI-Powered IT Operations

Builders vs Bystanders: The Forgotten Middle of the AI Revolution

Soon Seah Toh — Wed, 15 Apr 2026 00:04:22 +0000

Everyone's arguing about AI replacing jobs.

Here's what nobody's talking about.

The Binary Narrative We Need to Break

Yes, jobs will be displaced. That's not even debatable anymore. But here's the part that's being completely ignored in every boardroom conversation, every LinkedIn hot take, and every news headline:

Who's building the systems that do the replacing?

We've created this bizarre binary narrative — AI on one side, job loss on the other. Two extremes getting all the attention. But the massive middle? The actual work that needs to happen before a single job can be "automated away"? Crickets.

The Work Nobody Sees

Let me be blunt:

Before AI replaces your accounts payable team, someone has to build that system
Before AI handles your customer support, someone has to architect, integrate, test, and deploy it
Before AI writes your reports, someone has to connect it to your data, your workflows, your business logic
And after it's live? Someone has to maintain it. Monitor it. Fix it when it breaks at 2am

That's not a small job. That's an entire industry.

The Irony We're Missing

The irony is stunning. We're so busy panicking about jobs disappearing that we're completely blind to the massive wave of jobs being created — right now — in the space between "AI exists" and "AI runs your business."

23 Years of Building Enterprise Systems — Here's My Take

The people who will thrive aren't the ones hiding from AI. They're the ones who become the orchestrators. The builders. The ones who live on AI, work on AI, and understand its tools deeply enough to make it actually work in the real world.

Management? Your job isn't disappearing either — it's evolving. Your role is now strategy: figuring out WHERE AI fits, HOW to automate processes, and WHAT makes your business more efficient and profitable. That's harder than what you were doing before, not easier.

The Real Divide

The real divide won't be "humans vs AI."

It will be builders vs bystanders.

The builders — the ones constructing and maintaining these automation systems — they are the most critical and most overlooked people in this entire AI revolution. Without them, none of this works. No automation. No efficiency. No transformation. Nothing.

Stop Doom-Scrolling. Start Building.

The biggest opportunity of our generation isn't being replaced by AI. It's being the one who puts AI to work.

Stop doom-scrolling about job loss. Start building.

We're building exactly this at NetGain Systems — enterprise IT operations software that puts AI to work for real-world infrastructure monitoring and management.

See what we're building with V15 →

Written by Soon Seah Toh, CTO & Founder of NetGain Systems — 23 years building enterprise IT operations software.

Your Data Does Not Train Our AI. Here's Exactly How NetGain's Enterprise Deployments Work.

Soon Seah Toh — Sun, 12 Apr 2026 03:15:55 +0000

Every CTO we talk to asks the same question within the first five minutes.

"If your product uses AI, where does our data go?"

Fair question. It deserves a precise answer. Not a marketing slide. Not a vague "we take security seriously." A technical, verifiable, contractually backed answer that your security team can audit.

So here it is.

How Cloud Vista V15 Uses AI

Cloud Vista V15 is powered by Astra AI — five autonomous agents that handle anomaly detection, root cause analysis, predictive analytics, and autonomous remediation across your infrastructure.

These agents call large language models from Anthropic (Claude) and OpenAI (GPT) through their enterprise API tiers. This distinction matters enormously, because the enterprise API is a fundamentally different product from the consumer chatbot — with different data handling, different contractual terms, and different retention policies.

Most customer concerns originate from conflating the two. They are not the same.

This Applies to Everything We Build

Cloud Vista V15 is our flagship, but NetGain is actively developing custom AI solutions for customers across industries — automation platforms, intelligent workflows, predictive systems, and purpose-built agents tailored to specific business problems.

Every solution we build operates under the same security architecture described here. Same enterprise API tiers. Same contractual frameworks. Same deployment options. Whether it's Cloud Vista monitoring your infrastructure or a custom AI agent automating a business workflow, your data is handled under the same policies.

The Consumer Product vs Enterprise API — Why the Confusion

When people hear "ChatGPT" or "Claude," they picture the consumer chatbot. Consumer-tier products may use conversations to improve models under their default terms.

That is not what we use.

NetGain products call the enterprise API tier. Under enterprise API agreements:

Customer data is not used for model training
Retention is limited to short operational windows (and can be reduced to zero)
Data Processing Agreements are available for legal and compliance review

The distinction is not a marketing claim. It is a contractual commitment, backed by independent audit certifications.

Anthropic (Claude) — Enterprise API Data Handling

No model training on customer data. Under commercial API terms, inputs and outputs are not used to train or improve Claude models.
7-day default retention. Reduced from 30 days as of September 2025. Used for trust and safety screening only.
Zero Data Retention (ZDR) available. Inputs and outputs not stored beyond real-time abuse screening.
SOC 2 Type II audited. Independently examined for security, availability, and confidentiality.
ISO 27001 + ISO 42001 certified. ISO 42001 is the AI management system standard.
HIPAA BAA available. Native API is eligible under BAA, even without ZDR.
NIST 800-171r3 attestation available under NDA.

Reference: Anthropic Trust Center | Anthropic Privacy Center

OpenAI (GPT) — Enterprise API Data Handling

No model training on customer data. API platform, ChatGPT Enterprise, Business, and Edu data is not used for training. Opted out by default.
Configurable data retention, including zero retention.
SOC 2 Type II audited. Security, Availability, Confidentiality, and Privacy.
ISO 27001 + ISO 27701 certified.
HIPAA BAA available.
DPA available supporting GDPR, CCPA, HIPAA, and FERPA.

Reference: OpenAI Enterprise Privacy | OpenAI Security

Bring Your Own Subscription

NetGain does not require you to use our API keys. If your organisation already holds an Anthropic or OpenAI subscription, our solutions connect directly to your account.

Your billing, your controls. Manage usage limits and policies in your own dashboard.
Your DPA, your legal relationship. Direct contractual relationship with the AI provider.
Your audit trail. Full visibility into API usage in your own account.
No intermediary. Data flows directly from your instance to your subscription. NetGain does not proxy, intercept, or store this traffic.

The Actual Data Flow — Step by Step

The system detects an event — anomaly, metric spike, log pattern, topology change.
It constructs a prompt containing relevant operational data — metric names, values, timestamps, log excerpts. Scoped to operational telemetry.
Transmitted via encrypted HTTPS (TLS 1.2+/1.3) to the API endpoint.
The LLM processes and returns a response.
The response is consumed by the application. No persistent copy stored beyond the provider's stated retention window.

Under enterprise API terms: data is handled under agreements designed to prevent model training use and restrict retention to tightly controlled operational purposes.

Guardrails — How We Keep AI Controlled and Auditable

Model-Level Safety (Built Into Claude and GPT)

Anthropic's Constitutional AI: Safety constraints embedded during training. Constitutional Classifiers reduced jailbreak success rates from 86% to 4.4% in published testing. Hardcoded behavioural boundaries designed to resist override.

OpenAI's Safety Systems: Built-in content filtering, refusal mechanisms, configurable safety tiers for enterprise customers.

Application-Level Guardrails (Built by NetGain)

Input Controls:

Prompt sanitisation — system instructions architecturally separated from data payloads to reduce prompt injection risk (OWASP #1 LLM security concern)
PII detection and filtering — sensitive data patterns flagged or masked before reaching the AI
Scope constraints — each Astra AI agent confined to its specific domain

Output Controls:

Ground-truth validation — responses compared against actual monitoring data
Confidence scoring — low-confidence outputs surfaced with caveats
Output scanning — checked for sensitive content before presentation

Action Controls:

Configurable approval gates — you define what AI can act on autonomously vs. what requires human sign-off
Action scope limits — agents cannot improvise actions outside configured operations

Operational Controls:

Rate limiting — prevent runaway costs or API overuse
Full audit logging — every interaction logged for compliance and forensics
Kill switch — disable AI agents individually or collectively, immediate effect
Role-based access — integrates with your existing access control framework
Behaviour monitoring — drift detection with automatic alerting

The Guardrails Stack

Layer	What It Does	Who Controls It
Model Safety	Constitutional AI, content filtering, jailbreak resistance	Anthropic / OpenAI
Input Controls	Prompt construction, PII filtering, injection resistance	NetGain
Output Controls	Validation, confidence scoring, content scanning	NetGain
Action Gates	Human approval for high-impact actions	You (configurable)
Operational Controls	Audit logs, kill switch, RBAC, drift monitoring	You + NetGain

Defence in depth. Three independent layers, each catching what the others may not.

Cloud-in-Your-Tenant — Enhanced Isolation

Claude via AWS Bedrock: Data processed in your AWS account. FedRAMP High in GovCloud. DoD IL4/IL5 approved. Full AWS compliance suite.

Claude via Google Cloud Vertex AI: Data in Google Cloud perimeter. FedRAMP High. 10+ EU regions with regional endpoints.

GPT via Azure OpenAI: Data in your Azure tenant. ISO 42001 certified. 60+ regions. 100+ Azure certifications.

The Private LLM Option — Full Air-Gap

We support private deployment using open-source models on your own infrastructure. Your data does not leave your data centre.

We also believe in transparency about the real costs.

What It Actually Costs

Hardware: 8-GPU server (DGX H100): USD $300,000–$500,000 per node. Production requires 2+ nodes. $600,000–$1,000,000 in GPU hardware alone.

Infrastructure: Power upgrades ($50K–$200K/rack), cooling ($50K–$200K), InfiniBand networking ($20K–$100K), colocation ($5K–$15K/month).

The cost most underestimate — ongoing management:

MLOps staff (1.5–2 FTE): USD $260,000–$440,000/year. These roles take 3–6 months to fill.
Model lifecycle: Each major open-source model update = weeks of evaluation, testing, deployment work.
Security hardening: You own the full stack — CUDA, containers, OS, networking. Every component needs patching. Vulnerabilities at 2am are your problem.
Hardware failures: GPUs at sustained high power fail. Memory errors, thermal throttling, driver crashes.

3-Year Total Cost of Ownership

	Private LLM	Enterprise API
Year 1	$1,200,000 – $2,380,000	$36,000 – $96,000
Year 2–3	$1,050,000 – $2,020,000	$72,000 – $192,000
3-Year Total	$2,250,000 – $4,400,000	$108,000 – $288,000

The enterprise API tier delivers the security, contractual protection, and compliance posture most enterprises actually need — without the cost and complexity of private infrastructure.

If your regulatory constraints require on-premise, we can help implement it and connect you with managed hosting partners. But we recommend evaluating Bedrock, Vertex AI, or Azure OpenAI first.

The Six Tiers of Deployment

Tier	Best For
Enterprise API	Most enterprises. Fastest, strongest capability.
Your Own Subscription	Regulated industries needing direct legal control.
AWS Bedrock	Government, defence, financial services on AWS.
Google Vertex AI	GCP orgs, EU data residency.
Azure OpenAI	Azure orgs, broadest regional coverage.
Private LLM	Air-gapped, classified, sovereignty-mandated.

All tiers supported. Same product. Same agents. Same interface.

NetGain's Certifications

NetGain Systems is ISO/IEC 27001 certified covering development, deployment, and operation of Cloud Vista and all supporting infrastructure.

What We Do Not Do

We do not fine-tune public models on customer data.
We do not store prompts or AI responses beyond the operational session.
We do not share data between customers.
We do not use consumer-tier AI products. Enterprise API only.
We do not permit AI agents to execute destructive actions without customer-controlled approval gates.

The Summary for Your Security Team

Your data is processed through enterprise API tiers contractually restricted from model training use, independently audited under SOC 2 Type II and ISO 27001, with retention limited to 7 days or zero by agreement. Every AI interaction passes through layered guardrails — model safety, application controls, and customer-configurable approval gates. For additional isolation: AWS Bedrock, Google Vertex AI, or Azure OpenAI within your own tenant, or fully private on-premise. NetGain is ISO 27001 certified.

We built Cloud Vista V15 — and every AI solution we deliver — to provide full AI capability within a security framework designed to withstand enterprise scrutiny.

Because in enterprise IT, trust is not a feature you add. It is the foundation you build on.

For Data Processing Agreements, security architecture documentation, or deployment options: sales@netgain-systems.com

NetGain Systems — ISO 27001 Certified | AI-Powered Observability | Est. 2002

👉 netgain-systems.com/v15

My Best Co-Worker Runs on a Cron Tab

Soon Seah Toh — Fri, 10 Apr 2026 03:48:21 +0000

I have a co-worker who never sleeps, never takes leave, never forgets, and never complains about doing the same thing for the 500th time.

It's Claude. Running on four cron jobs on my MacBook.

Every morning before I open my laptop, this co-worker has already done a full day's work.

It scanned every Microsoft Teams channel and DM. Read every message. Figured out which ones need my attention — not just flagged them, but understood the context, urgency, and intent. Drafted replies I can send with one tap. Prioritized everything. High, medium, low.

The 3am Build Fix

Every 6 hours, it checked our CI/CD pipelines. When a build broke at 2am, it pulled the console log, diagnosed the error, checked out the branch, fixed the code, compiled it, pushed the fix, and triggered a rebuild. By the time I woke up, the build was green.

Every 15 minutes, Trello syncs. Daily, SSL certs auto-renew.

No product. No SaaS. No subscription. No vendor. Claude, a crontab with four lines, and Python scripts I wrote on weekends.

And it's a better co-worker than most humans I've worked with. I'm sorry, but it is.

The Math

A human doing this — triaging messages, monitoring builds, sorting emails — costs $3,000–5,000 a month. They'd still miss things. Still need sleep. Still scroll their phone at 3am. Still quit after 6 months because the work is soul-crushingly boring.

My Claude co-worker does it for the cost of an API call. And it gets better every week.

It Learns

Every week I tune what "actionable" means. It gets sharper at separating escalations from chatter. It learns which build errors are "just retry" versus "genuinely broken." It learns which emails are spam versus a real lead. My patterns, my priorities, my blind spots.

After months, it's become a shadow version of me. A digital CTO handling the 80% I used to waste my mornings on.

"But You Need to Be Technical"

Yes. You need to know what a cron job is and how to call an API. It's not no-code.

But it's not hard either. A cron job is one line: "run this script every hour." The script calls an API. Claude processes it. That's the entire architecture.

Two weekends. That's the gap between doing everything manually and having an AI co-worker that handles your morning. Two weekends and you're on steroids for the rest of your career.

"You're Letting AI Do Your Job"

No. I'm letting AI do the mundane job. The triaging. The sorting. The 3am build fixing.

So I can do the beautiful job. The thinking. The product decisions. The creative work. The customer conversations that actually move the needle.

My best co-worker doesn't have a face, a desk, or come to the Christmas party.

But it shows up at 3am. Every single night. Without being asked.

Find me a human who does that.

NetGain Systems — 23 years of enterprise software. Now with an AI co-worker on every cron tab.

👉 netgain-systems.com/v15

Anthropic Built an AI So Powerful They Refused to Release It. Here's What That Means for Enterprise Software.

Soon Seah Toh — Thu, 09 Apr 2026 11:18:46 +0000

Something happened this week that should make every software company stop and think.

Anthropic built their most powerful model ever. Claude Mythos. Then they looked at what it could do and said: "No. The world isn't ready for this."

Not a marketing stunt. A genuine safety decision. Only 12 organizations get access — AWS, Apple, Google, Microsoft, NVIDIA, and a handful of others — through an invitation-only program called Project Glasswing.

Let that sink in.

The AI That Found 1,000 Zero-Days Doesn't Sleep

It exploits vulnerabilities with 83% success rate on the first try. Thousands of previously unknown flaws across every major operating system and browser. In weeks.

We've crossed a line. AI is no longer just "a helpful tool." It's something that needs to be gated. Controlled. Released carefully.

So What Does This Mean for Enterprise Software?

It means the window is closing.

Every software company now falls into one of two categories: those who already embedded AI deeply into their platform, and those who are still "planning their AI strategy."

The second group is already behind. Not because the tech is hard. Because the gap between AI-native and AI-bolted-on is now so wide that customers can feel the difference in five minutes.

This Is Why We Built Astra AI

This is exactly why we built Astra AI into every layer of Cloud Vista V15 two years ago.

Not bolted on. Not a chatbot in the corner. Woven into the platform.

It watches every metric, every log, every trace. Simultaneously. It correlates anomalies the way Mythos correlates vulnerabilities — across your entire infrastructure, all at once.

Anomaly detection across metrics, logs, and traces
Root cause analysis — autonomous investigation
Autonomous remediation — with safety guardrails your team controls

Not a wrapper. The engine.

The Clock Is Ticking

The age of "we'll add AI later" is over.

If you're still treating AI as a feature on your roadmap instead of the foundation of your architecture, Mythos just showed you how fast that clock is ticking.

What's your take — are we entering the age of tiered AI where the most powerful models are invitation-only?

👉 netgain-systems.com/v15

Stop Building AI Into Your Product. Start Building Products With AI.

Soon Seah Toh — Thu, 02 Apr 2026 07:58:36 +0000

Everyone is chasing the wrong thing.

I watch company after company pour millions into embedding generative AI into their applications. RAG pipelines. Vector databases. Fine-tuned models. Prompt engineering teams. Guardrail frameworks. Data privacy reviews that take longer than the product development itself.

And after 18 months and a seven-figure budget, they have... a chatbot. A slightly smarter chatbot. That hallucinates 12% of the time and requires a legal review before every deployment.

Meanwhile, we just shipped a complete, custom-built automation platform for a customer in 6 days.

Not a chatbot. Not a wrapper around GPT. A real application. With real business logic. That does exactly what the customer asked for. No hallucination. No prompt injection risk. No data leaving the building. No AI ethics review needed.

How? We didn't put AI IN the product. We used AI to BUILD the product.

That distinction is worth billions. And almost nobody is talking about it.

The Dirty Secret Nobody Admits

I've been writing code for over 30 years. I've shipped products in C, Java, Python, JavaScript, and a dozen languages in between. I've led engineering teams, debugged production outages at 3am, and built monitoring systems that run critical infrastructure across Asia.

And I'm telling you — with zero ego protection — that AI writes better code than I do.

Not sometimes. Consistently. It handles edge cases I would have missed. It writes cleaner abstractions. It produces documentation I would have skipped. It refactors in ways that make me say "oh, that's smarter." And it does all of this in minutes, not days.

A 30-year veteran just admitted defeat to a machine. Are you paying attention?

Because here's what that means for your business: the cost of building custom software just collapsed by 95%. The thing that used to take a team of five engineers three months? One person and an AI coding agent can ship it in a week. Tested. Deployed. Working.

The In-Between That Everyone Is Missing

There are three approaches to AI right now:

Level 1: Ignore AI. Pretend it's not happening. Lose slowly, then all at once.

Level 3: Embed generative AI into your product. Build RAG. Fine-tune models. Deal with hallucination, data privacy, bias audits, and the regulatory minefield. Spend 18 months. Maybe it works.

Level 2: Use AI to build your product. No AI in the product itself. Just a brilliantly engineered, purpose-built application that solves the exact problem — built at 20-30x the speed of traditional development, with better code quality than human engineers produce.

Everyone is jumping from Level 1 to Level 3 and skipping Level 2 entirely.

That's insane.

Level 2 is where the immediate, massive, zero-risk value lives. No data privacy concerns — the AI never touches your customer data. No hallucination risk — there's no generative AI in the product. No bias audits — the application runs deterministic logic, just like every application before it. No regulatory uncertainty — it's just software.

But it was designed by an intelligence that understands your requirements at a level no human team can match, and built at a speed that makes traditional software development look like stone carving.

What This Looks Like In Practice

A customer tells us: "We need to consolidate data from four monitoring systems into one dashboard, with automated alerting when thresholds cross, and a weekly PDF report emailed to three managers."

Old world: 3 developers, 8 weeks, $120,000. Requirements drift halfway through. Delivered late. Half the features don't match what the customer actually wanted because things got lost in translation across 47 Jira tickets.

New world: We sit with the customer. Discuss requirements for an hour. Build it. Ship it. Iterate in real-time as the customer refines what they want. Done in days. The code is cleaner than what a human team would have produced. Every edge case handled. Every requirement met precisely because the iteration cycle is so fast that "misunderstood requirements" simply don't exist anymore — you just change it.

My estimate — and this is conservative — is that we're operating at 20 to 30 times the velocity of a traditional engineering team. Not because we're better engineers. Because AI is a better engineer than all of us, and the ones who admit that first will win.

The Privacy Argument Just Died

Half the companies I talk to say they can't use AI because of data privacy. "We can't send our data to OpenAI." "Our compliance team won't approve it."

Fine. I agree. Don't send your data anywhere.

But that argument is about putting AI IN your product. It has absolutely nothing to do with using AI to BUILD your product. The AI sees your requirements, not your customer data. It writes code, not processes sensitive information. The resulting application runs entirely on your infrastructure, touches zero external AI services, and is as private as any software you've ever deployed.

You've been using "data privacy" as an excuse to avoid AI entirely. But the excuse only applies to one approach — and it's not the one that gives you the fastest return.

Stop Overthinking. Start Building.

For the record — we're doing both. We're embedding generative AI into our products AND using AI to build them. The models are getting smarter every quarter. Data privacy solutions are maturing. The governance gaps are closing. And when those pieces fully click into place, the company that has been building with AI all along — shipping faster, iterating faster, learning faster — will have a product so far ahead that nobody else can catch up. That's the endgame. The ultimate killer product that nobody else can build — because they spent three years debating whether to start.

But here's the point most people are missing: you don't have to wait for Level 3 to get massive value from AI. Level 2 is sitting right there. No risk. No committee. No data privacy debate. Just AI building your product at 30x speed with better code quality than your engineering team produces.

The revolution isn't AI in your product. The revolution is AI building your product.

And if you're skipping Level 2 while chasing Level 3, you've already lost a year you're never getting back.

NetGain Systems — 23 years of enterprise software. Now building at 30x speed with AI coding agents.

Fire Your Automation Consultant. AI Just Replaced Them in 5 Minutes.

Soon Seah Toh — Fri, 27 Mar 2026 09:54:26 +0000

Your company paid someone $50,000 to spend three months interviewing your team, mapping your workflows, and producing a 60-page PDF that nobody read. The recommendation? "Automate your reporting." Wow. Revolutionary.

We did the same thing with AI. It took 5 minutes. And it was better.

Not marginally better. Embarrassingly better.

What Happened

We described a workflow. Plain English. "Our team spends 3 days every month pulling data from 4 systems, building charts in Excel, and emailing a PDF to management." That's it. No workshops. No stakeholder alignment sessions. No discovery sprints.

AI came back with: the exact integration points, a proposed architecture, an ROI estimate, a phased implementation plan, and three automation opportunities we hadn't even considered. In one conversation.

The $50,000 consultant missed two of those three.

The Bottleneck That No Longer Exists

This broke something in my brain. Because it means the entire "digital transformation consulting" industry is built on a bottleneck that no longer exists. The hard part was never building the automation. It was figuring out WHAT to automate. And humans are terrible at that — because we're blind to our own inefficiencies. We've been doing the same broken process for so long that it feels normal.

AI doesn't have that blindness. It has no loyalty to the status quo. It doesn't care that "we've always done it this way." It just looks at what you described and says: "Why is a human doing this?"

Self-Evolving AI Isn't a Research Paper. It's Tuesday Morning.

We've turned this into how we operate internally. Every single person on our team — engineers, ops, support — starts their week with one question: "What am I about to do manually that AI could do instead?"

It's not a productivity hack. It's a compounding engine. Every automation frees up time to find the next automation. We are literally using AI to evolve what AI does for us.

And the results are absurd. A 15-person company operating like a 50-person company. Not because we work harder. Because half the work simply doesn't exist anymore.

The Part That Should Make Consultants Nervous

We're building a portal where anyone — our partners, our customers, the general public — can sit down with AI, describe their daily work in plain language, and walk away with a full automation plan. Opportunities ranked. ROI calculated. Vision mapped. Implementation roadmap included.

No engagement letter. No discovery phase. No six-figure invoice. Just you, a conversation, and answers that used to take months.

The consulting industry won't tell you this, but the smartest thing you can do right now is ask AI what to automate — before someone else does it to your industry first.

Stay tuned.

NetGain Systems — 23 years of enterprise AI. Now using AI to find what AI should automate next.

You Already Know What to Automate. You're Just Not Asking.

Soon Seah Toh — Thu, 19 Mar 2026 12:53:43 +0000

"We don't know what to automate."

I hear this from nearly every company we talk to.

They know AI is powerful. They know there's opportunity. But they're paralyzed — afraid of picking the wrong thing, wasting money, or automating something that doesn't matter.

Here's the uncomfortable truth: the bottleneck isn't AI. It's not software. It's not budget. It's you.

The Answer Is Already Screaming at You

If a department head truly knows their job — really lives in it day to day — the answer to "what should we automate?" is already screaming at them. It's the thing they dread every Monday morning. The report they manually compile every month. The data they copy-paste between five systems. The checklist they've been doing the same way since 2014.

They don't need a consultant to tell them. They just need someone to ask:

"What takes up most of your team's time?"

"What do you do every week that feels like it should just... happen?"

Every operations manager we've sat with can answer those questions in under 30 seconds. The problem was never that they didn't know. The problem was that nobody asked — and even if they answered, nobody could build the solution fast enough to matter.

The Game Has Changed

That second part has changed. Dramatically.

Building software used to be the bottleneck. Months of specs, sprints, testing, deployment. By the time the solution shipped, the original problem had mutated into three new ones.

With AI-assisted development, what used to take 6 months now takes 6 weeks. Sometimes 6 days. That's not hype — that's what we're doing right now. Purpose-built automation portals, not generic chatbots. Real systems that replace real manual work.

A Real Example

A government agency manually prepares environmental monitoring charts every month — pulling data from multiple systems, plotting trends, annotating anomalies. Hours of work. Every month. For years.

We proposed a SaaS portal that auto-generates the reports, auto-detects anomalies, and delivers a PDF. Zero manual work.

That's one use case. The department head knew it was the problem immediately. He didn't need a strategy workshop or a digital transformation roadmap. He needed someone to ask the question and then actually build the thing.

Stop Overthinking

So the excuse "we don't know what to automate" doesn't hold up anymore. You do know. You've always known. You just didn't think anyone could solve it fast enough.

Now we can.

Stop overthinking. Start with the one thing your team complains about most. That's your first automation.

Automate SG — AI workflow automation for Singapore SMBs, powered by NetGain Systems' 20+ years of enterprise automation expertise.

Every Company Has a Process That Should Have Been Automated Years Ago

Soon Seah Toh — Tue, 17 Mar 2026 04:59:26 +0000

Every company has at least one process that should have been automated years ago.

We've spent 23 years building AI that monitors infrastructure — detecting anomalies, correlating root cause, automating remediation across thousands of servers and networks.

But here's what we've realized: the same AI capabilities that run a self-driving NOC can solve problems far beyond IT.

The Pattern We Keep Seeing

A government agency manually compiling environmental monitoring charts every month — pulling data from multiple systems, plotting trends, annotating anomalies. Hours of work. Every month. For years.

A hospital operations team drowning in compliance checklists that haven't changed in a decade — but still require someone to manually cross-reference five different systems.

A facilities management company producing the same weekly reports by copying data from spreadsheets into templates. Every. Single. Week.

These aren't technology problems. They're "nobody ever sat down with us and asked the right questions" problems.

The Two Questions That Unlock Everything

So that's what we've started doing. We sit with ops managers and department heads. We ask two simple questions:

"What takes up most of your team's time?"

"What do you do every week that feels like it should be automated?"

Most people can answer immediately. They've just never had someone who could actually build the solution.

How We Approach It

Step 1: We go talk to them. Sit down with ops managers, department heads — ask what takes up most of their team's time. Most people can answer that immediately.

Step 2: We identify the workflow. Could be monthly reporting, data consolidation from 5 systems, compliance checklists, document processing — whatever they describe. We map it out.

Step 3: We build a purpose-built SaaS portal — not a chatbot, not a generic AI wrapper — a system designed specifically for THEIR process. One-click automation. Monthly reports that generate themselves. Compliance checks that run while you sleep.

The Hard Part

The hard part isn't building it. We've been building AI systems for over two decades.

The hard part is finding the use cases. Every industry has them. Every company has them. They're hiding in plain sight — in the spreadsheets people dread opening, in the reports nobody wants to compile, in the processes everyone complains about but nobody fixes.

Looking for Partners

We're actively looking for BD partners who enjoy opening doors and having these conversations. People who can walk into a company, ask the right questions, and identify the workflow that's begging to be automated.

If that sounds like you — or if you're sitting on a painful manual process right now and thinking "that's us" — let's talk.

NetGain Systems — 23 years of enterprise AI, now solving automation challenges across every industry.

Saturday Night Thought: Full Self-Driving Is Coming to Your NOC.

Soon Seah Toh — Sat, 14 Mar 2026 10:09:00 +0000

The Self-Driving Analogy Nobody Is Making

Everyone talks about self-driving cars. Level 1 through Level 5. The idea that one day you get in, say where you want to go, and the car handles everything.

Nobody is having this conversation about IT operations. But they should be. Because it is happening faster than anyone expected.

Think about what a network operations center looks like today. Operators sit in front of dashboards. They watch metrics. They respond to alarms. They open tickets. They restart services. They allocate resources. They scale VMs. They disable compromised accounts. They adjust firewall rules. They do this all day, every day, shift after shift.

Now think about what an autonomous observability agent could do.

It reads every dashboard. Not one at a time like a human. All of them. Simultaneously. It ingests every metric, every log, every alarm, every topology relationship, every performance trend across your entire infrastructure. It does not blink. It does not get fatigued at 3 AM. It does not miss the correlation between rising CPU on one device and memory pressure on another because it was busy reading a different screen.

Then it thinks. It derives conclusions. Not just "CPU is high" but "CPU is high on this core router, memory is climbing on the connected distribution switch, and the last three times this pattern appeared it preceded an outage on this specific path." It connects the dots across time and topology in ways that would take a senior engineer an hour of digging.

Then it acts. Not recklessly. Through predefined, approved auto-actions. Create a ticket. Restart a service. Scale a VM. Allocate more resources. Remove temp files eating disk space. Disable a user account flagged by UEBA. Push a predefined firewall rule to contain a threat. Each action is something your team would have done anyway, just faster, and without waiting for a human to notice the problem first.

This is not science fiction. This is the trajectory we are on with Cloud Vista V15 and Astra AI.

What Happens to the Operators?

Same thing that happened to drivers when cruise control was invented. Nothing. They are still there. The car still needs a human. But the human's job changed. They stopped doing the repetitive, mechanical work and started focusing on the decisions that actually matter.

Senior engineers and architects are not going anywhere. Someone still needs to design the network. Someone still needs to make the judgment calls when the situation is genuinely novel. Someone still needs to decide strategy, evaluate risk, and approve the big changes. AI handles the routine. Humans handle the exceptions.

The junior operator's role does not disappear. It evolves. Instead of staring at dashboards and responding to alarms manually, they supervise the autonomous agent. They review its actions. They refine its playbooks. They become the person who trains the system to be better, not the person who does what the system could do faster.

Why 1-2 Years, Not 5

A year ago, I would have told you this was three to five years away. Today I am saying one to two years.

Two reasons.

First, we are building it. Astra AI inside our own observability platform already reads dashboards, derives insights, maps them to auto-actions, and executes. The pieces are not theoretical. They exist. They work. We can see exactly how full autonomy happens because we are making it happen step by step.

Second, the foundation models just leaped forward. GPT-4.5, Claude Opus 4.6, the latest generation of LLMs are not incremental upgrades. They reason better. They hold more context. They follow complex multi-step instructions with dramatically fewer errors. A year ago, getting an AI agent to reliably chain together "read dashboard, correlate anomalies, evaluate playbook, execute action, verify result" was fragile. Today these models handle that chain with a level of reliability that changes the engineering calculus completely. The reasoning gap that used to make full autonomy feel like a five-year problem just collapsed.

That sounds crazy. I know.

But if there is a customer out there willing to be a pioneer, or even a controlled test environment where we can prove it end to end, we are ready. Not in two years. Now.

The future of IT operations is not about replacing people. It is about an AI co-pilot that handles the 80% of routine work so your team can focus on the 20% that actually requires a human brain.

Full self-driving came to cars. Full autonomous observability is coming to your NOC.

The only question is whether you will be the one driving it, or watching someone else drive past.

"The best way to predict the future is to invent it." - Alan Kay

NetGain Systems builds enterprise observability software. Cloud Vista V15 with Astra AI is where we are building autonomous IT operations.