some ideas from @danshipper’s latest episode with @lennysan that are worth sitting with:
1/ “automation is a lie”
2/ “every automation creates a new thing to manage”
3/ “every agent needs a human”
4/ “agents need gardeners, not just users”
5/ “if nobody cares about the agent, it becomes useless”
6/ “a good agent is a maintained system, not just a smart model”
7/ “the future company may have one super-agent before everyone has a personal agent”
8/ “personal agents are coming, but company agents may come first”
9/ “the future of work may happen inside codex or claude co-work”
10/ “instead of putting ai into every tool, we may put every tool inside ai”
11/ “the browser inside the agent may matter more than the chatbot inside the app”
12/ “the next user of your product may be another agent”
13/ “software has to become legible to both humans and agents”
14/ “saas is not dead. saas may get more users than ever”
15/ “models make yesterday’s human competence cheap”
16/ “when competence gets cheap, taste becomes the advantage”
17/ “pm and design get stronger when building gets easier”
18/ “ai-generated writing will be everywhere, but only some of it will be useful”
19/ “slop is ai without taste”
20/ “leverage is ai with context, judgment, and direction”
the part i like most:
ai does not replace understanding
it rewards people who understand the work deeply enough to direct it
Have agents broken the unspoken contract we had with our databases? You know, human-authored apps running deterministic code and predictable queries?
@arpit_bhayani wrote a terrific post for how to create a defensively designed data layer ... https://t.co/8OOXrmkD6g
No developer tool is an island. Vertical integration among the editor, source repo, build system, CI/CD platform, runtimes, and support systems is important.
But it's hard to achieve, as Rebecca explains here: https://t.co/0vGsHhyRG7
Ok real talk here
One really bad thing about building agent systems with Claude Code or whatever...
Games aren't fun anymore.
I used to love playing a bit of Diablo IV every once in a while
And now it just doesn't compare to building
"[AI agents] are threatening to break the blood-brain barrier between the application layer and the OS layer."
Signal President Meredith Whittaker (@mer__edith) warns of "real danger" in agentic AI hype.
To be "magic genie bots" (concert booking etc.), they need root access to all your data: browser, cards, messages.
Cloud-processed & unencrypted: "privacy & security guarantees" at profound risk.
Here are my top five takeaways from the #QConLondon "Platforms, People & Process; Delivering Great Developer Experiences" track! 🚀
- Be intentional about creating a "Thinnest Viable Platform (TVP)." As @solidtubez repeatedly said, treat your platform as a product and balance the earning and spending of "trust credits" with your development teams. 🏗
- Balance standardisation and autonomy as the platform matures. Communicate and share technology choices, code libraries, and practices widely. Jemma Hussein Allen suggested using Tech Radars and Innersource 🗺
- You can't "shift left" everything without having a platform that supports and enables developers to deal with the increase in cognitive load. @aviranm showed us the gold standard for this with a case study from Wix 🧠
- Measure platform adoption rates, business impact, and developer experience. Ana Petkovska provided a fantastic overview of the journey her platform team took to reduce the number of developer support tickets 🎯
- "Developer experience creates experienced developers." This quote from @andyburgin will stick with me. He showed the value of devex and engineering enablement and highlighted the need to continually evolve your efforts as the teams grow and mature 👨🏫
It was a genuine privilege to be able to host the track, and I want to thank all of the speakers, as well as the attendees, organisers, sponsors, and volunteers 👏
Software is never done until it’s out of production.
Instead of seeking a “definition of done,”
let’s define “enough.”
Enough on this feature, time to shift our focus again.
This forms an interesting "law" or "rule" of work.
The more things that are going to be started soon, and the more things in flight at once, the more I'm going to be interrupted.
@JustinKitagawa Given that you showed an overall net promoter score, did you ask one single overall NPS question? Or one for each of these 9 statements (and calculated the overall score)?
@JustinKitagawa Did you ask “Agree / Disagree” from 1-10, or use the classic format of “How likely are you to recommend X to a friend or colleague?” If the latter, how did you phrase this?
@natpryce Tried Act for GitHub Actions?
It’s not official (I wish), but it works for running almost all things supported in GitHub Actions workflows: https://t.co/SwqXuo2JDQ
@ThaiWoodHere Any chance for an RSS feed? 😍
I hvae been subscribing for quite a while but have been noting that over time, email hasn't felt like the medium I personally use for casual reading/learning.