Saiful Aziz

I'm a Cloud and IT Consultant. All opinion are my own. Talks about #secdevops #multicloud #datacenters #hybridcloud #cloudinfrastructure #costoptimization

[Jensen Huang @JensenHuang]

For my first post, I’m sharing a letter @NVIDIA signed on why open models matter.

AI will transform every industry, power every company, and be built by every country.

Open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty.

The world needs both frontier closed models and frontier open models.

https://t.co/AUKzoQ5Ikb

[OpenAI @OpenAI]

Introducing Daybreak: frontier AI for cyber defenders.

Daybreak brings together the most capable OpenAI models, Codex, and our security partners to accelerate cyber defense and continuously secure software.

A step toward a future where security teams can move at the speed defense demands.

[klöss @kloss_xyz]

let me explain what Anthropic just did

they built an AI model so good at finding security vulnerabilities that they have refused to release it

meet Claude Mythos

→ it’s Anthropic’s newest frontier model and it’s not available to the public. not because it’s not ready. because it’s too dangerous

→ Mythos found tens of thousands of zero day vulnerabilities across every major operating system and web browser… many of them 1 to 2 decades old. for context… Opus 4.6 found about 500. Mythos found tens of thousands

→ it found vulnerabilities in the Linux kernel. a 27 year old vulnerability in OpenBSD. a 16 year old vulnerability in FFmpeg
→ it doesn’t just find bugs. it writes the exploits too. that’s the part that scared them

→ so instead of releasing it… Anthropic has created Project Glasswing. a cybersecurity initiative where they hand picked 40+ companies to use Mythos for defense only

→ the partner list reads like a who’s who of tech… Amazon, Apple, Microsoft, Google, Nvidia, Broadcom, Cisco, CrowdStrike, Palo Alto Networks, JPMorgan, the Linux Foundation

→ Anthropic is giving up to $100 million in usage credits to these partners and $4 million to open source security organizations

→ they’re briefing CISA and the Commerce Department on how to handle this

→ the benchmarks are truly insane… Mythos hit 77.8% on SWE-bench Pro where Opus 4.6 scored 53.4%. hit 93.9% on SWE-bench Verified where Opus 4.6 scored 80.8%

→ Anthropic’s head of frontier red team said this is “the first time a model is this good that we decided to approach release in a very different way”

this is the first time an AI company has held back a model because it was too capable

not too expensive. not too slow.

too dangerous

and instead of locking it in a vault they weaponized it for defense and gave it to the companies that run the internet

that’s either the most responsible thing an AI company has ever done… or the scariest

only time will tell

[Saiful Aziz @SaifulAziz]

@davemorin @steipete Nice 👍

[Saiful Aziz @SaifulAziz]

🧵 How I cut a client's Lambda bill from $10K → $2,500/month:

1/ They were running 300 Lambda instances 24/7.
(Yes, serverless doesn't mean "always on")

2/ The mistake? No concurrency limits.
Every request spun up a new instance.
Cost explosion.

3/ The fix: Provisioned concurrency = average request rate.
Simple math. Massive savings.

4/ Result: 75% cost reduction.
$7,500/month back in their pocket.

Most Lambda waste comes from treating "serverless" like traditional servers.

If Lambda feels “unpredictably expensive,”
your concurrency model probably is.

Any other use case or want to share tips and tricks? Drop a 👇

[Saiful Aziz @SaifulAziz]

"The obstacle on the path becomes the path", wrote Marcus Aurelius. The rain outside, though heavy, mirrors the work within — there will always be friction, delays, or cascades beyond your control.

What remains yours is the choice to begin anyway. You have projects worth building, skills worth sharpening, and another full day to apply yourself.

Start where you are. Use what you have. Do what you can. The rest is noise.

[Saiful Aziz @SaifulAziz]

I spent 6 years managing AWS infrastructure. Here are 7 security mistakes that block AI projects in regulated industries:

1/ Sending data to public AI APIs "It's just for testing." Until compliance finds out. Use PrivateLink. Now.

2/ Default VPC endpoints Bedrock works out of the box. But your data leaves your network. Configure VPC endpoints first.

3/ Ignoring model governance Who accessed what? When? Audit trails aren't optional in finance/healthcare.

4/ Overly permissive IAM "Allow * on *" for AI services? That's a breach waiting to happen. Least privilege. Always.

5/ No data classification Not all data can go to AI. Tag first. Classify second. Route accordingly.

6/ Skipping encryption at rest Your prompts contain PII. Encrypt them. It's one checkbox.

7/ Single-region deployment AI services have outages. Multi-region isn't paranoia. It's business continuity.

The shift: Security isn't a blocker. It's an enabler. When done right, it speeds up AI adoption.

Which one hit hardest? Let me know 👇

You can learn more here to understand more related with security best practices for agentic AI systems on AWS: https://t.co/k8jqvjUMgX

#AWS #Bedrock #AISecurity #PrivateLink #CloudSecurity #FinServ #Healthcare #DevOps #TechTwitter

[Saiful Aziz @SaifulAziz]

Unpopular opinion: AI won't replace your FinOps team. It'll make the bad ones obsolete.

Here's why most people get this wrong:

1/ AI needs clean data: Most AWS environments are tagging disasters. AI can't optimize what it can't categorize.

2/ Context matters: AI can tell you Reserved Instances save money. It can't tell you that 3-year commit kills your architecture flexibility.

3/ Execution is human: AI suggests. Engineers decide. Finance approves. The loop needs people.

The winners? Teams using AI for analysis, humans for strategy.

Agree or Disagree? Tell me why 👇

#AWS #FinOps #GenAI #CloudCost #DevOps #TechTwitter #CloudComputing #AIOps

[Saiful Aziz @SaifulAziz]

Amazon OpenSearch Service now supports Graviton4-based instances — boosting performance & efficiency for search and analytics workloads.

What this means:
✅ Faster query performance
✅ Better throughput for analytics
✅ Lower cost per workload with efficient CPUs
✅ Scales intelligently with high-performance infrastructure

A great win for anyone running search, logs, metrics, or real-time analytics at scale.

🔗 https://t.co/mcVYZjWNQ1

#AWS #AmazonOpenSearch #Graviton4 #CloudPerformance #SearchAnalytics #CostOptimization #CloudInfrastructure #DevOps #BigData

Follow me for simple, practical cloud & search insights 🚀

[Saiful Aziz @SaifulAziz]

Big update for AI builders!

AWS just added reinforcement fine-tuning to Amazon Bedrock with OpenAI-compatible APIs.

This makes it easier to:
✅ Improve model accuracy with feedback (not massive labels)
✅ Fine-tune models faster, end-to-end
✅ Build domain-specific, production-ready GenAI
✅ Keep data secure inside AWS
✅ Deploy immediately using familiar APIs

This is a big step toward practical, cost-efficient GenAI at scale.

🔗 https://t.co/dF6P5i1EJT
#GenerativeAI #AmazonBedrock #ReinforcementFineTuning #OpenAICompatible #AWS #MachineLearning #CloudInnovation #GenAI

Follow me for real-world cloud & AI insights

[Saiful Aziz @SaifulAziz]

@JulianGoldieSEO No its not free, I play with macmini + openclaw + kimi 2.5 since yesterday and I need to pay to use its API key (.ai, .cn or Kimi code subscription.

[Saiful Aziz @SaifulAziz]

5️⃣ Hot take:

AWS isn’t chasing “best model” headlines — it’s betting that enterprises care more about control, cost, and capacity than raw model IQ.

Open weights + self-hosting + ARM economics > closed per-token APIs.

This wins long-term. Agree or disagree?

[Saiful Aziz @SaifulAziz]

Top 5 AWS Highlights This Week
Nova inference beyond Bedrock, new EC2 M8azn, Amazon’s $200B AI bet, and Bedrock open weights.
If you run AWS at scale — this week mattered 🧵👇

[Saiful Aziz @SaifulAziz]

4️⃣ Bedrock expands open-weight foundation models

Not just API access anymore — downloadable model weights reduce vendor lock-in and unlock air-gapped & regulated deployments.

AWS stance is clear:
API for speed. Weights for control.
🔗 https://t.co/pA9LzRcHpn

[OpenClaw🦞 @openclaw]

🦞 OpenClaw 2026.2.12 is out!

🔥 GLM-5 + MiniMax M2.5
💬 IRC channel — your bot fits right in with the old guard
🛡️ 40+ security fixes
📦 Custom provider onboarding, compaction improvements & more

Your agent called. It wants an upgrade.
https://t.co/rezHViFGhU

[Shubham Saboo @Saboo_Shubham_]

https://t.co/gupiw10kRV

[VraserX e/acc @VraserX]

I completely agree with Sam Altman; within 2 to 3 years, most knowledge work won’t be about doing the work anymore. It will be about directing AI agents.

The real skill will shift from execution to orchestration.
Knowing what to ask. Structuring problems clearly. Making high-leverage decisions while fleets of agents handle the heavy lifting.

Managers of intelligence instead of producers of output.