As AI agents become more capable of accomplishing real-world work at human-equivalent competencies, we will need to reorganize our division of labor. Some changes will disrupt how a job role is performed. Other changes will change what job roles are needed or how many people are needed to do the work. AI agents that automate tasks can be seen as tools for accelerating a person's work. A person can do more and they can do it faster with AI tools. However, the job role remains firmly the human's responsibility, because their know-how, judgement, and accountability are irreplaceable. AI agents have no moral or ethical grounding. An AI agent cannot be held liable. The buck cannot stop at an AI agent. A human job role exists for those purposes. As work is accelerated, we will see responsibilities roll up and concentrate into human job roles that have expanded scope of control. Work bottlenecks are primarily due to the slow speed of knowledge transfer, collaboration, and consensus-building between humans. The more that multi-person collaboration can be replaced by solo work, the more efficient thr work becomes. Job roles will become more multi-disciplinary, as responsibilities concentrate into solo work for improved efficiency. If coders no longer write the code and they no longer read the code, the job of engineering software shifts from coding to overseeing requirements, design, quality assurance, and building machines that do those activities along with doing the coding. This disruption and displacement is emotionally taxing. Egos get bruised. It's very difficult for highly-paid professionals to accept that their knowledge, skills, and competencies can largely be automated. Their experience remains valuable, but it needs to be applied differently. Instead of toiling as they used to, the knowledge, skills, and expertise need to be directed into training, building, and running machines that do the job role they used to do by hand.

I got fed up with /implement pushing tiny commits to create a pull request containing slices of one or two tests with barely any functionality. I asked why its slices were so tiny. The agent answered that it was minimizing risk. I never asked it to minimize risk. I am now asking it to accept higher risk. Be ambitious. Aspire to be like an X games competitor jumping off the edge of the world in a squirrel suit. Fly and be free. Yolo!

Because software architecture and design are all about trade-offs, you will start to hear me repeat the following: Why are you chasing a 140+ IQ frontier model to perform a 100 IQ use case?

In the long term, I believe that tokens will be commoditized so that 99% of usage will target 100 IQ inference targeting 100 IQ human audiences. Only the remaining 1% of tokens will need 140+ IQ inference for all other use cases.

Hey @mattpocockuk, /implement-spec doesn't seem to choose by default to prioritize the success path, which enhances features functionally to deliver the most value, ahead of failure handling and edge cases. I have to explicitly guide it to do so. Nice tweak to the skill?

For projects with containerized deployments, after a while you will notice very poor hygiene with regard to containers and images that accumulate. The process of determining which images should be retained and which ones should be deleted is tedious. There ought to be a skill. More sustainably, there ought to be a way to specify rules for hygiene, so the normal life cycle operations prevent this from becoming a problem.

My wife and I went from paying over $300/mo for television packages from the local cable provider to now only paying for Netflix and Hulu. I also used to buy ESPN+ along with a NHL package. When a subset of games needed subscriptions to other providers, I got fed up and cancelled everything. If I cannot purchase a single subscription for all games of interest to me, I won't pay at all.

I need to purchase a track pad or mouse jiggler to prevent the MacBook from blanking the screen and disconnecting the VPN, so that long-running tasks in Codex can run uninterrupted without needing me to babysit only to manually jiggle the pointer.

What is going on at Anthropic with high profile contributors to Claude Code exiting before the anticipated IPO? Seems like these individuals are foregoing a ton of personal wealth without an obvious opportunity as motivation. Their public statements remain positive toward Ant, so there is no obvious disenchantment. Very curious.

How many generations of humans have existed since homo sapiens split from our common ancestor with other homo species? «Roughly 20,000–30,000 generations, most plausibly around 20,000–25,000» -Grok If we take science seriously and we take the story of genesis seriously without pitting one against the other, we must interpret these as telling the same history with consistency. Using different styles and words, but two perspectives witnessing the same truth in reality.

We need a software revolution around how to maintain the secrecy of secrets. AI enabled exploits of security vulnerabilities will continue until we get this under control. The biggest weakness in the face of attacks is the lack of adequate protection for secrets.

Don't trust me, bro. Rely on machine verification of evidence.

You really expect me to create a 12 character password for encrypting a file, and be able to remember it into the future, after never needing to use it on a regular basis? You are crazy. After forgetting my Obsidian vault sync password, I am well aware of how impossible it is to manage these kinds of secrets.

Tibo is responsive, timely, and polite to the toxic and rabid customer base that inhabits this platform. The majority of replies to his posts are rude and disrespectful. People are so shitty.