mercurialsolo

IMO visualization is such a critical part of explanation - we have created an entire suite of tools to visualize - presentations, graphs, formatted docs, websites, svgs, interactive blocks, apps.

To that front - model providers generating visualization elements (i,e taking over the tools for interaction and visualization) is just a natural next step in value capture.

Anthropic launched docs, presentations, sites and am sure they have sheets next up their alley. Apps are artifacts. Tool forming is a natural model adjacency.

I see reasoning models being used to change the way to draft assignments and formulate teaching. Accelerating the learning gradient for students now requires a better understanding of how they learn and working with their motivators.

E.g. if students learn better through gameplay - can assignments become like game levels?

Rather than building for verifiable outputs - build assignments that gauge process and help elicit what they have learnt not what they have produced.

One of the engineers I know is building this product called SkillEd for just this. Lemme know if you need an invite

We just need better specialized models for these. Writing is as much a process of thinking as of communicating the idea. Current generation of outputs eliminate the process of thinking

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.