Not sure if it’s just me, despite having weekly usage left hitting 5 hour limit has been a real blocker. Would be great to have the ability to move past 5 hour limit especially if there is weekly usage left. Or nonetheless an opportunity to improve user experience as it’s clearly been a friction. Surprisingly, haven’t faced this with Claude Code
It’s important to pause and process how quickly has agentic traffic grown. Unlike the era where a builder would think whether to prioritize mobile experience or not, making your product usable by agents is becoming increasingly important.
Will be interesting to hear @bgurley's perspective in one of the upcoming podcasts or posts. Feel that such policies are a good tailwind for open-source models as well as for companies building their own models. This is definitely a big moment in this whole AI era we are living in.
@keyur19198@levie Agreed. Another analogy - printing press didn’t shrink writing it exploded it. Pamphleteers, novelists, journalists, scientists. Cheaper production expanded who could participate.
AI is doing the same to knowledge work. Excited to see what new roles open up when the floor drops?
@keyur19198@levie Agreed. Another analogy - printing press didn’t shrink writing it exploded it. Pamphleteers, novelists, journalists, scientists. Cheaper production expanded who could participate.
AI is doing the same to knowledge work. Excited to see what new roles open up when the floor drops?
@levie being a thought leader in this space, laid out the perspective very well for leaders and investors.
My POV based on customer research calls- Previously a lot of non-technical people were constrained by technical resourcing or skills to use certain tools. That barrier is slowly fading. For example at @Opendoor, as a PM there were problems that Ops team faced which required engineering to build something. However, now those operators can build their own tooling without being constrained on another team having to unblock them. This kind of an unlock creates a flywheel effect where the net unit of value delivered by individual increases which leads to more hiring in software and certain non software roles.
We are still in early phases and the demand for everything that will lead to AI enablement may it be humans or software will continue to increase aka Jeavon’s paradox!
This is very much true for the quality of skill files people created to instruct agents. Through user research at KramaAI, I have seen the struggle of getting agents to execute in skill files that do not execute the way you would want to.
Why does it matter?
With @OpenAI 's GPT 5.5 launch today, this year feels like the year where computer use will continue to gain adoption, whereby the modal capabilities for GPT 5.5 to use tools and integration jumped to 55.6%(massive jump from ~38%, 6 months ago). Super impressive!
@OpenAI just launched workspace agents in ChatGPT.
The hard part isn't shipping the agent. It's describing what it should actually do.
Krama AI captures how you already work and turns it into a skill file your agent can use. You don't describe the agent. We already have.
@keyur19198@Reuters While Meta installs keystroke trackers to train AI that does your job for you. Krama is building the opposite. AI that meets me in my flow and makes me sharper. My data. My terms. My win.
The problem is that the top 10% that has figured out how to use AI effectively in an organization cannot easily elevate the next top 10%.
It’s a knowledge lock in problem
I know this topic is top of mind for Fortune 500 and in general companies with large workforce. Any leader planning AI Transformation, this is a very well-written article and framing. Also recommend reading part 1 if you haven't.
@garrytan Belated happy birthday @garrytan! Gstack feels like working in a company with all these cadences but all managed by agents. Continuing to stay on top of it for my startup KramaAI.
A female sea lion leaping onto a fishing boat while carrying her pup to escape a pursuing orca, showcasing instinctive maternal defense🤍
📍 Seattle in November 2025,