Technically Monthly (July 2025)
How AI models “think” and “reason”, why code takes so long to ship, how developers choose cloud infrastructure, and more.
Hello there, illustrious Technically readers. This is the fourth ever Technically Monthly, and like that Honda Civic you’ve been driving for what feels like 20 years, it’s small but gets the job done. We’ve got for you, in no particular order:
3 new posts on Technically 2.0: all about code reviews (why your features take forever), how AI models can think and reason, and a breakdown of cloud infra
Technically for Teams upgrades
An oldie but important term to know: the multi-tenant architecture
New on Technically
Breakdown: The Cloud Infrastructure Market
Available as a sneak peek to paid subscribers on Substack, and now in its permanent home in the Analyzing Software Companies knowledge base.
Where does software actually run—and why are there so many vendors selling what looks like the same thing? Understanding the cloud infrastructure market can feel like a herculean task.
This post dives into the full stack of cloud infrastructure: IaaS, PaaS, and SaaS; what actually gets deployed (frontends, backends, databases, analytics, etc); and why developers choose one vendor over another.
If you’re going deep on companies who play in this universe – or you’ve ever wondered how dev teams actually make cloud infrastructure choices in the wild – this one’s for you.
Developers Hate This One Thing (All About Code Reviews)
Available free to all readers, thanks to CodeRabbit, who helps expedite the code review process with AI.
If you’ve been waiting on your developers to build a new feature for what feels like years…the culprit might be code reviews. The code is written, but stuck in some draconian review / feedback process that rarely ends up helping at all.
What are code reviews, and why do they take so long? How are they changing with more and more code being written by AI? And how do AI tools like CodeRabbit and Sweep try to speed them up?
We dig into the sociology of code review (who comments, who merges, and when), as well as emerging strategies that aim to automate the tedious parts.
If your PRs are rotting like week-old lettuce, this one’s for you.
All About AI Reasoning Models
Available as a sneak peek to paid subscribers on Substack, and now in its (free) permanent home in the AI, it’s not that complicated knowledge base.
OpenAI’s new o3 model and DeepSeek’s R1 are part of a growing trend: LLMs that don’t just predict the next word—they reason through problems and think step by step.
This post explores the rise of “reasoning models,” which aim to replicate humanlike chains of thought. We explain what makes these models different, how they’re trained, and why this shift matters for real-world applications (from math problems to multi-step search).
It’s also an early look at where GenAI might be headed next: not just faster or cheaper, but smarter.
Technically for Teams
After last month’s teaser about Technically for Teams, we’ve decided to open the curtain (or something) and share more about how we help teams get smarter: we’ve got plans for you to check out if you run a team of analysts or a team of sellers.
A couple of small but important updates:
To support those of you without a Google Workspace account, Technically now supports sign-in with email. Sometimes the little things, like actually letting your customers log in to your product, make a big difference.
The Analyzing Software Products knowledge base (home of Technically’s company + category-level breakdowns) is now available as a monthly subscription (rather than just annual). Try it out for a bit before going in for the whole year.
Technically has a lot of stuff to read in it…it can be daunting. Last week we started doing “Intro to Technically” sessions for teams who are aiming to become more technical, which have been a lot of fun.
From the Universe: Multi-tenant architectures
Let’s close with a favorite term from the Technically universe of technical concepts, the multi-tenant architecture.
When you work in a technical company, you’ll hear how much easier + cheaper it is to support customers in a multi-tenant (shared) environment vs single-tenant (dedicated) environment:
Multi-tenant simply means shared with someone else, like having a roommate in an apartment. In the infrastructure context, it refers to one physical server that has software from multiple different people (or companies) running on it. Which could be hundreds of roommates depending on the size of the server.
Multi-tenancy is central to why cloud was such a revolution in the 2010s: it let providers like AWS rent space on a server without having to allocate the entire server to one person. Unless you pay extra, most of the cloud services you use are probably sharing hardware with someone else.
Coming up next month
Thanks for reading this edition of Technically Monthly. On deck for next month:
The long-awaited GPU post (yes, it's finally happening)
A clear-eyed answer to a surprisingly slippery question: What is JavaScript?
And a breakdown of what generative AI really is—and what it definitely isn’t