Made some use of Claude Fable 5 while it was available for a short window.
Built a small game : https://t.co/TfRoejta4M
(Opus 4.8 did heavy lifting on the feedback after fable access was gone)
@sarahookr Maxwell food centre: you can try Mr. Appam (https://t.co/iGcMyc1tbu), also many options for Chinese food and Vietnamese food.
More options on Indian Food: Murugan Idli, MTR, Kailash Parbat
@khyatimaloo and jungles were really pristine as some of the sites were never open for public till 2010s. Nope just did this hike+camping for 2 days 1 night.
Try giving full read only access to AWS using cli to claude code along with any md/architecture files you have already prepared. Tell it to check architecture on it's own and create it's own context /update your architectural context. Then Ask it to go through the cost report on AWS, and ask for final output/report you wanted. It will eat lot of tokens but you should get a lot better output. You can cross-check/reuse the architectural context that it generated as well.
@omarsar0 I recently helped my wife setup a telegram chatbot connected to openclaw agents (running on hardened VPS) where she can just assign some repetitive marekting tasks like interlinking etc via chat. All she needs to do is sharing access to files, excel task sheets etc to a service account email, and then talk to agent via telegram. While it is still far from perfect I can already see huge appeal for non-tech people and use-cases.
Personally I still used Claude Code to plan/develop this whole thing :)
Running a company is just context engineering internally.
Now that skill has even more value in the agentic world. Us tech founders have been doing reps to prepare for this.
There has always been a push towards abstraction in telling machines what to do. If we look at machine instructions per human-input: Growth has been linear with assembly, c/c++, python. It's just that it has gone parabolic now with AI and could go to infinity once required human input is 0. We all feel nostalgic about the time we started participating
@divgarg It has to be as simple as a whatsapp message or a voice note. WhatsApp has already penetrated into almost all village households in India. The AI experience should not be any different than just another contact or a number to talk to for the mass public
https://t.co/8sf56DeP92
Urban company usage must have grown among expats in SG. In the condo over the weekend, It is a common sight to see UC employees waiting for entry to lift/main gate. Used their services twice and it was very reliable. I guess Target market is young expats and couples who do not have kids or elders at home, otherwise full time house help is still more affordable and norm in SG.
@aarthir I think the entry would be to solve a particular use case/automation where the context graph creation is a necessity, rather than directly selling this whole thing as a solution to enterprises
, and then evolve from there
Amazing read @JayaGup10
While trying to automate a freight audit for an enterprise customer we realised the most complex part was not parsing documents correctly etc. But it was to capture hundreds of rules that define the audit process. Some of those were documented well, some just existed as human knowledge and exceptions were pretty common. The thing that moved the needle was to map this whole thing out from multiple customer calls, excels and decision traces. Ultimately it's a graph that is running the automated system. Of course AI helps in creating this context graph
I'm not sure why there's a need to justify the 100 Cr price tag for everyoneβit may only be worth it for those willing to pay that amount. For reference, this was the view from a flat over the last few months, which would have cost under 10 Cr in central Singapore, where clean air and sunshine are the norm, not the exception.