Issue #594

Essential Reading For Engineering Leaders

Friday 25th February’s issue is presented by Paragon

Building function tools for AI agents is time-consuming and hard to optimize.

  1. There are hundreds of possible actions you may need to automate across hundreds of actions across apps like CRMs, ticketing, & productivity platforms.

  2. You need to figure out the 3rd-party API’s auth, understand the right endpoints to hit / chain, write agent-optimized tool descriptions, & build in monitoring.

Instead of having engineering build & optimize each function tool, use ActionKit.

It’s a single API to give your AI agent 1000+ integration actions such as CREATE_SALESFORCE_CONTACT, DRAFT_GMAIL_EMAIL, GET_CALENDAR_AVAILABILITY, SEARCH_ZENDESK_TICKET.

— Ted Neward

tl;dr: “There's a whole host of mistakes that companies often fall prey to with respect to those they have leading teams, and I thought it a good idea to collect them into one place, under the umbrella heading of "manager antipatterns”.”

Leadership Management

— Michał Poczwardowski

tl;dr: “Press pause and take your time to help yourself by: (1) Detaching from emotions that blur your judgement at the moment. (2) Making sure that you can check for blind spots. (3) Getting a new perspective. People may demand answers immediately, and the pressure can be high, but they rarely argue with rules. You can say: "I have a rule that I never answer immediately.”

Leadership Management

tl;dr: “Why doesn’t my AI agent call the right tool?” For AI agents that automate work across 3rd-party tools to be adopted, tool calling and tool use accuracy is critical. Join this webinar, where our dev advocate will cover the basics around AI agent tool calling as well as evaluations we ran to uncover optimizing tool calling performance. Register here - you’ll receive the recording and slides afterward, even if you can’t make it.

Promoted by Paragon

BestPractices AI Management

— John Ousterhout, Robert Martin

tl;dr: The document that features a dialogue between Robert "Uncle Bob" Martin and John Ousterhout discussing their differing software design philosophies. They debate three main topics: method length, comments and test-driven development.

BestPractices

"Learning never exhausts the mind."

— Leonardo Da Vinci

— Ben Howdle

tl;dr: Over the years, I've spotted a pattern: all engineers exist on a spectrum between speed and accuracy. Some lean towards speed, optimizing for fast iteration and progress, while others prioritize accuracy, ensuring long-term maintainability and scalability. Neither end of the spectrum is "better" than the other, but knowing where you sit — and understanding what kind of engineer your company actually needs — can be the difference between thriving in a role or feeling completely out of sync.

CareerAdvice

— Zack Proser

tl;dr: “Every time a device connects to your server, it broadcasts a wealth of information through its browser. Some of these signals are obvious, while others are subtle technical artifacts of how browsers and hardware work together.” Zack breaks down what servers can see and how to mitigate bad actors.

Promoted by WorkOS

BestPractices Security Management

— Scott Chacon

tl;dr: From the co-founder of GitHub: “In this post, I’ll go through some of the perhaps obscure Git config settings that I have personally globally enabled and go into them to explain what they do and why they should probably be the default settings.”

Git

— Brian Kihoon Lee

tl;dr: "I don’t think I’m making any outlandish claims here - merely simple extrapolations of trends that I think have robust foundations, with a dash of history. And yet the conclusions are surprising. The world of 2030 will be as unrecognizable to us as the world of cellphones today was in 2010, and as the world of the connected Internet was in 2000."

— Simon Tatham

tl;dr: “macOS is fortunate to have access to the huge arsenal of standard Unix tools. There are also a good number of macOS-specific command-line utilities that provide unique macOS functionality. To view the full documentation for any of these commands, run man <command>.”

CommandLine

Arroyo: Distributed stream processing engine.

BeatCode: Live LeetCode battle with abilities.

KubeVPN: Cloud native dev environment.

Tach: Python tool to enforce dependencies & interfaces.

Upfetch: Advanced fetch client builder.


How did you like this issue of Pointer?

1 = Didn't enjoy it all // 5 = Really enjoyed it
1  |  2  |  3  |  4  |  5

Login or Subscribe to participate in polls.