Building in the AI revolution and sharing what I learn about staying creative, productive and valuable. Previously Atlassian, Trello, Ubiquiti Networks.
Los Angeles, CA
Joined February 2009
- Tweets5K
- Following1.6K
- Followers864
- Likes4.2K
I CANNOT wait for models to be fast enough that the alpha isn't in how many ways you can split your attention. I HATE HATE HATE mental multi-threading.
The promise of operating at a higher abstraction level that doesn't require so much attention thrashing is there is pockets but the reality just hasn't fully arrived yet.
Agent: "You need to tromble the woogles on the gorbenshtein platform..."
Me: "Can you just do that for me?"
So many times a day...
Travis Fischer retweeted
Ok, this is big!
MCP now defines an extension specifically for discovering and loading Agent Skills from MCP servers
(Save this, It’s worth reading how it works)
The flow is pretty simple:
- MCP Server
- Discover available Skills
- Get Skill metadata
- Load the SKILL.md only when needed
This is a huge step for better context window usage
Instead of loading everything upfront, the agent can discover Skills and load only the one it needs from the MCP
Great progress from the Model Context Protocol team! 👏
github.com/modelcontextproto…
They have been for a minute. And also... the hard part about making games is NOT the implementation. Way more so than business software. The models are not particularly good at figuring out "fun" or "interesting" yet.
That said, a lot of latent game design geniuses are now equipped to build. So ultimately, I co-sign.
Sol is not coping well with the Astra launch...
Response just now:
> Wait need correct path. I inadvertently wrote malformed. Need fix. Ensure actual path. [normalization-service.ts?] use correct. Need continue. We cannot edit after final. Need produce correct. Let's redo full response carefully.
Cloud coding agents seem so obviously like "The Way" it's suprising to me how not first class that paradigm is in the OpenAI and Anthropic product stacks.
This class of take, isn't gonna age well.
Yes, obviously you can just give an agent an gmail address an that will mostly work in most cases. And while I can't tell you the exact set of features that you would want email for agents to have today, I can tell you that there are a large and ever increasing set of reasons agent <> human.
This fact will become more relevant the more agents can do and agent-native / agent-specific products will find huge tailwinds.
Building an agent that quietly slides a $20 bill across the agentic counter while making loaded eye contact when requesting said reservation.
I can count the number of times I've actually used an apps' "built-in agent" (beyond the obligatory "demo" run) on exactly 0 hands. I know I'm not every user but I think most existing SaaS apps don't need both.
Travis Fischer retweeted
there's so much left to be done in the application layer for agents - the success of @bot shows that
we're barely scratching the surface
This is (directionally) the way.
(1) Building with agents is a highly adjacent/overlapped but new/different skill from building the old fashioned way.
(2) You learn a surprising amount watching someone work with an agent. I would argue it's actually much higher signal than watching them hand roll code because they have a lot more leverage at their disposal and you can see if/how they use it said leverage.
We're using a new method to hire engineers at @HelloUntangle.
I think this playbook will become the standard for hiring technical talent in the age of agents.
Here's how it works:
1. We ask candidates to submit a screen recording of their whole screen, shipping a new feature on a current app they work on.
Note: This is usually 45 minutes of video - no editing or speeding up - we want to see them thinking, prompting and problem solving with the agent - we also want to see their devops skills.
2. We watch these videos and pick a subset to go through Round 2.
3. Round 2 is giving them a seat on @DevinAI so they have secure access to our repo.
4. We ask them to ship a new real-world feature and get the PR merge-ready.
5. We pay them a fixed fee to do this, and they have 16 hours to complete the work.
6. We watch the videos, review at all the Devin threads, have an agent do the same, and then choose which one to hire.
Note: We don't have any meetings with the folks before any of these things. It's all about the work and the results.