Chandan

Building AI cloud infrastructure • Founder @TorontoAIMeetup (10K+) • Neo Cloud • Posts on LLMs, agentic AI, GPU economics

[Chandan @kchandanAI]

Just wrapped up @TorontoAIMeetup event in Bay Area at @Rakuten SFO office. Learned a lot and looking forward to come back soon

[Chandan @kchandanAI]

All of sudden firing reason is no longer AI.

[Polymarket @Polymarket]

JUST IN: Uber to cut 23% of its People decision, claiming the layoffs are “unrelated to AI”

[Chandan @kchandanAI]

This is wild. Worse victim of this would be smaller neo cloud players who are paying almost double price for these servers due to Memory shortage, only ti realize by the time their H/W hits market, they are already underwater

[Thierry from arvy 🇨🇭 @ThierryBorgeat]

The price to rent an Nvidia H200 just collapsed from $7/hr to $4/hr in three weeks.

A -40% drop in the cost of the single most strategic asset in tech.

When the underlying commodity that powers your entire thesis loses 40% of its value in a month, that usually means one of two things: supply finally caught up, or demand was never as deep as the headlines said.

Either way, somebody is selling.

So why is the AI trade still pricing in scarcity?

[Chandan @kchandanAI]

I don’t think agent is eating any entry level job at all, especially in large corporations. On the other hand tech companies might be little different. Vast majority of layoffs are done due to reason unrelated to AI.

Tech companies firing due to AI but not because of agents but because of AI infra spent

[Chandan @kchandanAI]

Every 𝐢𝐧-𝐩𝐞𝐫𝐬𝐨𝐧 interaction teaches you more than just asking something to a chatbot.

Being a tech professional, it’s very rare that you get to cross to the other side and look at your job and skills from a different perspective.

Yesterday’s @TorontoAIMeetup panel discussion taught us a lot. As always, our rock star moderator, Mohit Rajhans, made it look so easy that we didn’t even realize how quickly the time passed.

𝐅𝐫𝐨𝐦 𝐚 𝐍𝐞𝐰𝐬/𝐌𝐞𝐝𝐢𝐚 𝐏𝐞𝐫𝐬𝐩𝐞𝐜𝐭𝐢𝐯𝐞

Ian Knopke — No matter what tool you use to curate your news article, if you (a human) publish it, you own the content and its risks/rewards.

𝐅𝐫𝐨𝐦 𝐚𝐧 𝐈𝐏/𝐋𝐞𝐠𝐚𝐥 𝐒𝐭𝐚𝐧𝐝𝐩𝐨𝐢𝐧𝐭

Stephano Salani — No matter what tool you use to generate the code or vibe code, you should be able to understand and explain what it does and its consequences.

𝐅𝐫𝐨𝐦 𝐄𝐧𝐭𝐞𝐫𝐩𝐫𝐢𝐬𝐞 𝐚𝐝𝐨𝐩𝐭𝐢𝐨𝐧

Chandan K. - When you allow an AI agent or tool to be embedded into your enterprise ecosystem, treat it just as you would treat a new employee. Don’t give away all your access rights right away. Test, validate, and verify in a dev instance/sandbox before you even let it touch production systems.

Special Thanks for Anny Ditmars and Dipchand LLP for hosting the event and allow our community member to enrich with knowledge and sharing.

[Chandan @kchandanAI]

@BenjDicken @Hamzeml I’m glad to work both on MOMs and MOPS. Worked for MariaDB SkySQL launch ( Covid era pre IPO) and Postgres mostly through out my career till date. Now mostly redshift and Lakebase

[Chandan @kchandanAI]

I am getting the same feeling too. Lot of enterprise and business looking for copilot based workflow automation.

Just last week, I hand copilot workshop for Marketing team of a large company, where we taught them how they could be copilot Studio to do lot of stuff without going to IT

[Sam Badawi @Sam_Badawi]

Satya Nadella says AI is becoming the new “Lean” for knowledge work at $MSFT, helping companies automate processes and reduce waste.

Meanwhile, $NOW and $CRM are both well off their lows. Maybe AI isn’t killing software after all… maybe it’s making the best platforms even stronger.

[Chandan @kchandanAI]

Microsoft numbers are even bigger

[Chandan @kchandanAI]

Here is the simple chart of Meta Employee growth over past decade

Take a look into Covid era hiring, ofcourse Market cap and revenue is completely different story

[Chandan @kchandanAI]

We are looking for startups to show/demo their Tech products for our upcoming July demo day.

If your solution is built for Developers, Tech users, please post it here.

You will have opportunity to reach over 10K tech professionals across GTA and US east coast.

@

[Chandan @kchandanAI]

@amitisinvesting They are preparing for potential rate hike

[Chandan @kchandanAI]

@unusual_whales Its already game over for many folks. First round of victims are developers like us

[Chandan @kchandanAI]

I used AI to build my AI resume, only to be rejected by AI

[Chandan @kchandanAI]

In my experience, vibe coding does not completely replaces developer. It makes developer more productive and more busy. Because many times they would overcommit thinking they got "vibe power", only to realize that Dev is now the QA, DevOps combined. I think in long term ( when AI dust settles), it would be back to normal, more code, more bugs, more feature and higher code velocity.

[Chandan @kchandanAI]

𝐀𝐈 𝐌𝐚𝐝𝐞 𝐂𝐨𝐝𝐢𝐧𝐠 𝐄𝐚𝐬𝐲. 𝐓𝐡𝐞 𝐄𝐧𝐠𝐢𝐧𝐞𝐞𝐫𝐢𝐧𝐠 𝐉𝐨𝐛 𝐆𝐨𝐭 𝐇𝐚𝐫𝐝𝐞𝐫

Back in 2019 I wrote a Medium post on the journey to becoming a DevOps engineer. It ranked #1 on Google for "𝐝𝐞𝐯𝐨𝐩𝐬 𝐞𝐧𝐠𝐢𝐧𝐞𝐞𝐫 𝐣𝐨𝐮𝐫𝐧𝐞𝐲" for years. The tools in it are dated. The principles aren't — but the job has gotten harder, not easier.

𝐇𝐞𝐫𝐞'𝐬 𝐰𝐡𝐚𝐭 𝐜𝐡𝐚𝐧𝐠𝐞𝐝.

Writing code used to be the bottleneck. In 2026, it isn't. A second-year CS student with Cursor and Claude Code can ship a working CRUD app in an afternoon. Generating a passable Terraform module is a 30-second prompt. Translating a Jira ticket into a working PR is a feature, not a skill.

So if writing code is cheap, what does an engineer get paid for?

𝐓𝐡𝐫𝐞𝐞 𝐭𝐡𝐢𝐧𝐠𝐬, 𝐚𝐥𝐥 𝐦𝐨𝐯𝐞𝐝 𝐮𝐩 𝐭𝐡𝐞 𝐬𝐭𝐚𝐜𝐤:

𝟏. 𝐕𝐚𝐥𝐢𝐝𝐚𝐭𝐢𝐧𝐠 𝐭𝐡𝐚𝐭 𝐭𝐡𝐞 𝐜𝐨𝐝𝐞 𝐝𝐨𝐞𝐬 𝐰𝐡𝐚𝐭 𝐭𝐡𝐞 𝐬𝐩𝐞𝐜 𝐚𝐜𝐭𝐮𝐚𝐥𝐥𝐲 𝐫𝐞𝐪𝐮𝐢𝐫𝐞𝐬. LLMs produce plausible code that compiles. They do not produce code that handles your edge cases, your tenancy model, your idempotency requirements, or your hidden assumptions. Validation is now the job.

𝟐. 𝐃𝐞𝐬𝐢𝐠𝐧𝐢𝐧𝐠 𝐬𝐲𝐬𝐭𝐞𝐦𝐬 𝐭𝐡𝐚𝐭 𝐟𝐢𝐭 𝐭𝐨𝐠𝐞𝐭𝐡𝐞𝐫. 𝐀𝐧𝐲𝐨𝐧𝐞 𝐜𝐚𝐧 𝐬𝐜𝐚𝐟𝐟𝐨𝐥𝐝 𝐚 𝐬𝐞𝐫𝐯𝐢𝐜𝐞. Few can decide whether it should be a service at all, where the boundaries go, what the retry semantics need to be, and which failure mode you're willing to accept. System design didn't get automated — it got more important.

𝟑. 𝐎𝐰𝐧𝐢𝐧𝐠 𝐩𝐫𝐨𝐝𝐮𝐜𝐭𝐢𝐨𝐧 𝐰𝐡𝐞𝐧 𝐭𝐡𝐞 𝐚𝐠𝐞𝐧𝐭'𝐬 𝐥𝐨𝐠𝐢𝐜 𝐛𝐥𝐞𝐰 𝐮𝐩 𝐚𝐭 𝟐 𝐀𝐌. Generated code fails in generated ways. The pager still rings. The person who can read someone else's (or some model's) code, find the bug, and ship the fix is the person who keeps a job.

The "vibe coding to demo" trap is real. The path from working demo to working production is exactly where AI-generated code falls apart, and that gap is widening, not closing.

𝐖𝐡𝐚𝐭 𝐭𝐨 𝐢𝐧𝐯𝐞𝐬𝐭 𝐢𝐧 𝐢𝐟 𝐲𝐨𝐮'𝐫𝐞 𝐚 𝐃𝐞𝐯𝐎𝐩𝐬 𝐞𝐧𝐠𝐢𝐧𝐞𝐞𝐫 𝐢𝐧 𝟐𝟎𝟐𝟔:

● System design fundamentals — load balancers, queues, caching, idempotency, eventual consistency
● Property-based testing and fuzz testing — the test cases LLMs won't think of
● Reading code 10× more than writing it
Code review as your core skill, not a chore
● Eval-driven development for any code that touches an LLM
● Tools are cheap. Judgment is expensive.

The career value of the engineer who can tell when an AI is wrong has never been higher.

[Chandan @kchandanAI]

Great move !

[Mark Carney @MarkJCarney]

It’s time to build.

[Chandan @kchandanAI]

Rate Odd lots Miss. Tracker Supply. May be we need another episode on this to understand exactly what happened?
@TheStalwart @tracyalloway

[Chandan @kchandanAI]

@refsrc Don’t think they did much work other than following the dollar on AI trade. TSMC == Taiwan.

These kind of ill informed report ask Claude to create for you 😆

[Chandan @kchandanAI]

@gregorykennedy A single prompt can create a whole ebook. Don’t know what will happen to folks on fivr who used to sell service for customization of ebook.

[Chandan @kchandanAI]

@ThierryBorgeat It’s no secret that it’s K shape. It has been going on for sometime in Canada. Not sure about US

[Chandan @kchandanAI]

Some securities never recover @ArkkDaily

While rest of the market is Up.