Everyone wants an agent that can do the work at their company. An agent that specializes in their problem domain. The trouble is there's no manual on how to generically build an agent that can solve their problems. This presentation on cyber security presents a framework that can work under any business domain. Crawl Walk, then Run.
I've seen it work and am excited to go through the process again.
I gave 15 popular LLM models the PID-5 personality assessment at least 10 times each. 9 out of 15 LLMs had clinically significant personality findings. I'm not saying we should diagnose those LLMs with conditions from the DSM, but it definitely feels concerning.
For Junior Developers it is the best of times and it is the worst of times...
Architects used to draw blueprints by hand. Then there was CAD. Old-timers complained about 3D modeling and how the next generation was getting ruined. How they weren't learning to think about buildings. How the profession would be wrecked by this new technology.
Nope.
Vibe coding is sitting back and letting the LLM drive. It is accepting that I don't care what my assembly looks like, why should I care what my javascript looks like? It is trusting an LLM more than the hundreds of library authors who's code I run every npm install.
Interesting conference vibes. There were a few really interesting people doing interesting things. There were also a whole raft of really ambitious people wanting to tell you about the product/company they were launching.
AWS is incredibly powerful. Their products are awesome. Their ability to execute is unparallelled.
Their ability to fumble the ball so completely in some areas is also breath taking. Tagging is one of those features.
Dear Users,
We are shutting down Cyclic.
Effective immediately we are turning off signups by new users.
In two weeks, May 10th, we will begin turning off free tier usage.
We will begin turning off all usage May 31.
In software value is produced by features running in production. Let’s assume a team delivers on average one feature per unit of time. We will call it team velocity. We can plot their feature delivery curve.
When building a startup I rarely have the time to build software in the way I think of as correct. In these situations I find particular pleasure when simple solutions can be "good enough". Best yet is when one of these solutions holds up to increasing numbers of users, employees and use cases. One pattern that manages all of these is using Slack as a searchable event stream DB.