mercurialsolo

This whole thing reeks of AI slop writing - wtf allowed this to be on the main page

Not your weights not your brain. Owning your own action and decision model is super important as these models emulate more of our decisions, thinking and learning. Built claudectl - a local brain for coding agents https://github.com/mercurialsolo/claudectl

The best design is invisible - most (web)sites are designed for text based reading / watching - primary modality. Maybe we will see more inspired design - with voice, video or agent scanners using which one can talk to an agent via an assistant

every wrapper either gets acquired or stays long enough to be a zombie startup

15 mins into this - Built this to identify the fraudsters https://github.com/mercurialsolo/realstars

We should do a hall of shame!

Cost of signalling is way lesser than the cost of verification.

Stars are like for developers? and you have a bunch of creators now entering the arena. what did you expect?

Why not just do event based triggers e.g. register (web)hooks instead of schedules time based triggers. Have a mechanism to listen to an event and then run some flow - analyze, plan, execute, feedback

For the past few days have been thinking on the future of open source in the agent era - a lot of more coding agents and source code gets generated by agentic tools. Small essay on how I see open source being reimagined for the new user - agents.

This whole post sounds - LLM generated

Codex has now caught up to Claude Opus and this is a defensive move by Anthropic

The slop is everywhere

I wonder if we need to "bitter lesson" this - aren't general techniques gonna outperform any constitution / laws which seem more rule based?

Clickhouse needs observability models to be more useful to agent run infra

Software markets are in a thermopylae kind of pass - every inch matters now

A perspective on where opportunities lie in the markets today.

I had tried manus and never could find a use-case for them that worked for me

1. Insanely overpriced versus over deep research products 2. Deep research has increasingly become a feature in most other products 3. They shot themselves in the foot by sharing very limited usage credits, in the initial wave of DR products pretty much everything was free - ChatGPT, Claude, Pplx, Deepseek. they rolled this back later and added a free credit tier but by then the hype had moved off.

TBF 1. Their post synthesis, formatting abilities were better than others 2. Their initial launch was "hypey" - lots of waitlist based access.

But I had seen somewhere they mention they had hit $100mn in revenue - M&A also signals that DR is increasingly a feature of the labs. And labs missing an assistant will probably buy a well distributed one

claude code works tho

shopify.com

As is supabase

Trust goes offline. The value shift is happening in realtime as more in-person events, offline meetups take more value over digital communications and meetups. You can forge in the digital space but the real you is in-person.

Not just Management, I also see the auth APIs down - anyone else?

Skills again seem more like making MCPs more accessible for everyday users. It will again need to see evolution - things that are missing - containers for skills (you will need beyond a folder for sharing a toolchain with skills) - the orchestration and picking up of the skill is left to the LLM (the synthesis is done via the context shared in the skill - this feels a very sub-optimal pattern at the moment - low control)

Few others like versioning, access control for tools which are missing.

Real world skills come not just from practice but are opionated workflows built with specific toolchains too.

IMO, this is half-ass engineered at the moment

robots in the warehouse talking to robots on the streets talking to car robots. starting of robot swarms

da hype is real

The way this is headed - I also see a burgeoning class of tools emerging. MCP servers, Skill managers, Sub-Agent builders. Feels like the patterns and protocols need more explainability to how they synthesize into a practical dev (extension) toolkit which is useful across multiple surfaces e.g. chat vs coding vs media gen.

One sharp contrast though I see between OpenAI and Anthropic is the product extensions are built around their flagship products.

OpenAI ships extensions for ChatGPT - that feed more to plug into the consumer experience. Anthropic ships extensions (made for builders) into ClaudeCode - feel more DX.

Sub agents, mcp, skills - wonder how are they supposed to interact with each other?

Feels like fair bit of overlap here. It's ok to proceed in a direction where you are upgrading the spec and enabling claude wth additional capabilities. But one can pretty much use any of these approaches and end up with the same capability for an agent.

Right now feels like a ux upgrade from mcp where you need a json but instead can use a markdown in a file / folder and provide multi-modal inputs.

This is a new amazon marketplace with trusted reviews. IMO having worked in commerce for a while - this opens up new avenues for merchants to work on gaming reccomendations and the trust quotient delivered.

It also opens up - ads - subscriptions - transactions

based revenue lines for OpenAI

/rewind was a must needed upgrade for the agent.

- Need better memory management and controls (especially across multi-repos) - /upgrade needs better management