Ryan Smith

STATE OF THE PRODUCT JOB MARKET IN EARLY 2026 In spite of the headlines about layoffs and AI taking jobs, we’re actually seeing a lot of promising signs in tech hiring, and some interesting new trends: 1. PM openings are at the highest levels we’ve seen in over three years 2. AI hasn’t slowed the demand for software engineers (at least not yet) 3. AI roles in general are absolutely exploding 4. Design roles have plateaued 5. The Bay Area is increasing in importance 6. Remote work opportunities continue to decline 7. Despite ongoing layoffs, the overall number of tech jobs continues to grow More in 🧵

@nozmen This is great. More of this. I see the AI adoption and usage, and to pull apart behaviour is insightful. I'm working on a PR verification against the original plan.

Core contributors and occasional contributors don't use coding agents the same way. Paper from the University of Saskatchewan. 9,427 agentic PRs on GitHub across Claude Code, Copilot, Cursor, and OpenAI Codex. 1,701 developers, 1,391 repos. They split developers into core (top 20% by contribution) and occasional (everyone else). Both groups use agents at similar rates, but some occasional contributors rely on them much more (mean 6.08 PRs vs 3.73). Core developers mostly use agents for docs and testing. Occasional contributors spread tasks evenly across bug fixes, features, docs, and tests. 74.1% of agentic PRs get accepted without any modification. The gap is in verification. Occasional contributors merge without running CI checks nearly twice as often (19.1% vs 11.2%). Core developers' PRs also land on main/master more (85.8% vs 77.8%). The authors' concern: occasional contributors submitting agent PRs without engaging deeply with the code, which could slow down long-term learning.

@ryancarson @DevinAI I'm in pruning hell right now - cleaning up branches, worktrees, docker images, ... docs that have drift, ... with great empathy for all the dev managers out there.

Paid $500 for @DevinAI - liking it so far. You can tell this team is much further than other agent labs when it comes to being truly remote-first. Very mature, advanced tooling and it just works across all surfaces (iPhone, Slack, Browser, GitHub, Linear). I was tired of trying to hand-connect everything with a custom setup of Open Inspect + Codex + Linear. When you look at your hourly effective rate, it stops making sense trying to hand-build all this stuff. Nvm the all the maintenance hours you need to put in. I'll keep using Devin 100% for the next week and report back. So far, my PR shipping velocity is higher than before - so that's good obv.

@luongnv89 linear.app - which has an MCP, though I build a robust skill that handles a ton of the project management work github.com/wrsmith108/lin…

@W_RyanSmith what is Linear? i use /github-issue-creator skill also for that, it also works pretty ok

if you often "waste" your usage when the resetting time coming > you can prepare tasks before hand, and best place to do it > right in your project's issue on Github. i am using /github-issue-creator to automate the process > support multiple input: screenshots, error log, email, etc. > follow your template/ or fallback to a standard one > duplicate detection > approval gated > batch creation just need to tell what is the problem, issues will be created, with relevant context, ready for agent to pick-up and work on. install via @vercel : > npx skills add github.com/luongnv89/skil… --skill github-issue-creator install via asm: > asm install github:luongnv89/skills:skills/github-issue-creator get it to tweak: github.com/luongnv89/skil…

@BrandonWaselnuk that's what Claude AI is for on your phone 📱🔥💂‍♂️

i didn't use all my claude max tokens today cause I'm tourist-maxxing in London. can't believe I didn't live up to my potential today

@luongnv89 I'm curious for your review of Skillsmith, Loung, as I'm also working to separate the dependency issues from the trigger word issues and minimize silent fails by skills. Love to chat.

nightmare of ai agents is here. which skill should be installed, then if multiple installed, which skill to use when the prompt say "review my changes" i am trying to solve that problem

@GiladPeleg @trq212 have you got into autoresearch? It does require training data, then you can benchmark and show success / fail data on a skill, with improvements against the benchmark

@W_RyanSmith @trq212 None of the above :) more like a meta "is this skill effective/efficient/adheres to best practice"

@patrickbo11e @pranav_vinodrg Awesome!

I built a user-submitted directory of all the niche groups, clubs, and communities happening in Vancouver. It's all free and open-source. vancouvercommunity.org Find a philosophy club, dinner club, chess party, hiking group, or whatever you are looking for in the city.

@trq212 IF you're still new and figuring out Skills, we have an April intensive over three weeks to go from writing 1 skill to chaining sets of skills together in workflows! maven.com/agenticaipm/pr… #MADSKILLZ

attention all Canadian builders (in SF and in 🍁) -we're hosting a 🇨🇦 demo night at our new SF HQ -we're picking 2 Canadian builders to fly out for free if you want a free trip to sf comment what you're working on, i'll dm you march 26th nice people, nice food and nice demos

Who's into an intensive on Agent Skills in April? @trq212 making skills beautiful @AnthropicAI maven.com/agenticaipm/pr… #ClaudeCode @Cursor @OpenAI @antigravity friendly

@aabyzov @trq212 I'm into quality. Let's Jam.

Although it contains an advanced decision engine, I do believe most efforts are concentrated around creating skills with the best secure, reusable way. Nobody is talking about the quality of skills. We need to start dedicating time writing unit and integration tests for skills. Managing test cases and evals could be as easy as running this command npx vskill studio I'd be happy to help with this.

@GiladPeleg @trq212 Like a publishing command to clean up, or CI/CD, or security?

@trq212 Looks like another category is needed for meta skill analysis/lint/review

@eatpraydiehard @trq212 Framework below. I love Skill Invoked Parallel Agents skillsmith.app/blog/agent-ski…

@trq212 do you guys use mostly subagents w/ skills or skills that launch subagents? feels like there is a lot of overlap between the two patterns

@__rmrf @trq212 Varlock.dev or compose a 1Password skill, keeping the keys and secrets out of agents hands, rendering them at runtime

@trq212 How to handle sensitive keys in Skills though?

@travelingmunchi @trq212 Dependency intelligence is a good start, then cull overlapping skills, or refactor into many small skills composed into workflows. Dependency Intelligence: How Skillsmith Infers What Your Skills Need | Skillsmith Blog share.google/IOqCA3fKPxGkzl…

@trq212 How can we better manage should at an Enterprise level. Multiple divisions and departments within divisions. All have overlapping and colliding skills. Guidance I have seen is about having an "internal registry" and process to evaluate and approve skills. Need tooling pls!

@aabyzov @trq212 try Skillsmith optimize which will refactor the skill with a thin dispatcher and sub-skills in the same folder. skillsmith.app/blog/agent-ski…

@trq212 Also have you considered lazy loading of skills? When you realize that something that you have in the marketplace would boost the current project dev pace or maybe even a third party that you know is verified and reliable...

@trq212 Love skills. Skillsmith has a whole set of tools now for authors - create, optimize, secure, even a dependency graph! skillsmith.app for the CLI and MCP