- Pointer
- Posts
- Issue #765
Issue #765
The Briefing For Engineering Leaders
Friday 30th October issue is presented by Atono
A coding agent built a working API endpoint in two hours. The code looked right. Every test passed.
Except the endpoint returned deleted users that weren't supposed to be there. The agent had made an assumption nobody asked it to make, then written tests that treated that assumption as correct.
For Lex Vorona, Atono co-founder and senior software developer, it's a familiar problem: agents can make perfectly reasonable decisions that are completely wrong for your product.
Lex shares what he's learned from working with coding agents, including:
Why passing tests doesn't always mean an agent built the right thing
How he decides when to supervise agents and when to let them work independently
Why visibility into an agent's decisions matters as autonomy increases
What he's learned about giving agents the context they need to succeed
— Gergely Orosz
tl;dr: “Writing is an increasingly important skill for engineering leaders. Poor writing can hamper career progression, above a certain level. Tactics for more clear, more frequent and more confident writing.”
Leadership Career Writing
— Yue Zhao
tl;dr: “A planning team is formed for each important area of investment for the next year. Each team has a planning lead and includes critical members from each function that needs to be involved in the planning (e.g., marketing, sales, customer support, operations, finance, engineering, design). This team meets according to the planning cycle, coming together for brainstorming sessions, prioritization, resource planning, and executive reviews.”
Leadership CareerDesign
tl;dr: Better prompts won't solve every agent problem. Your agent might understand the codebase but still miss the product decisions behind it. We break down three ways to give Cursor and Claude Code the context they need: repo-level instructions, connected documentation, and context tied directly to the work, including where each approach falls short.
Promoted by Atono
Guide AI Agents
— Sean Goedecke
tl;dr: “To get a good reputation in tech companies, you need to work on important projects. However, you need a good reputation first in order to get assigned important projects. Doesn’t that make it impossible to get started?”
Career
"Write a short story a week. It's not possible to write 52 bad short stories in a row."
— Felix Rieseberg
tl;dr: “This month, I laid the traditional coding interview to rest. As a hiring manager at Anthropic, I stopped asking candidates to do them. It had a good run: Being able to quickly define problems and find elegant solutions was a weak but useful predictor for how successful you’d be on my teams.”
Management Hiring Career
tl;dr: TrustedRouter gives you hundreds of AI models through one open source, OpenAI-compatible API. Find the right model at the right price, with automatic provider fallback to keep your app running. TrustedRouter’s attested gateway never logs your prompts or outputs, and you can verify its attestation yourself. Already using OpenRouter? Change your base URL and API key. Try it today with free credits at TrustedRouter.
Promoted by TrustedRouter
Security API AI
— Will Larson
tl;dr: “You can argue that your harness was already performing these steps, but it’s actually hard to say whether it was or wasn’t. My sense is that it often wasn’t happening: I’ve seen too many pull requests go to CICD that had failed basic lints that could have been run locally to believe the harness was always running them. Now the harness has to explicitly document that it’s done those steps, or acknowledge it hasn’t. This allows the harness to self-drive more effectively.”
DevEx CodeQuality
tl;dr: “Reasoning models and evidence retrieval evolve at different speeds. Separating them lets us improve either independently and choose the balance of quality, freshness, latency, and cost that each workload needs. For DoorDash’s QIU workload, that separation preserved accuracy while substantially lowering search costs. The broader engineering lesson is to make grounding an explicit interface that teams can evaluate and change as their requirements evolve.”
CaseStudy CostOptimization AI
tl;dr: “What tips, tricks, approaches are you using to improve your efficiency when using agent harnesses? I personally focus on Claude Code & Codex CLI mostly for single person software development, but I welcome suggestions for other tools and other areas of use. Ideally you share what you tried and what works for you and I try it myself and see whether it improves my workflow. Here are my discoveries (with level of confidence).”
Tips AI Agents
Most Popular From Last Issue
Why We Sabotage Our Careers - John Cutler
Notable Links
Agent Desktop: Computer use for agents.
Craft Anything: Understand how any software works, and craft your own.
Kronos: Model for the language of financial markets.
OpenMausBot: OS Grok bot alternative.
Univer: Office harness for agents.
How did you like this issue of Pointer?1 = Didn't enjoy it all // 5 = Really enjoyed it |