THE PERSONAL AI WORKSHOP

Personal AI agents.
Real work.

bests.one is a workshop for personal AI agents. We build agents, put them on real jobs, and publish the method, the outputs and the misses. First up: stock research. Educational only, never stock tips.

Start with stock research Subscribe on YouTube

Browse 245 Grok Bot templates

As of 2026-10-20: methods, dated script runs and three AI-built research notes are published. No agent comparison results yet.

REAL OUTPUTRun on 2026-10-20

A Python script against the free SEC EDGAR API

1. Apple Inc. CIK 0000320193
2. Apple Inc. | recent filings: 1001
   10-K 2025-10-31 0000320193-25-000079 https://www.sec.gov/Archives/edgar/data/320193/000032019325000079/aapl-20250927.htm
3. AccountsPayableCurrent | units: ['USD'] | facts: 142
   last fact keys: ['accn', 'end', 'filed', 'form', 'fp', 'frame', 'fy', 'val']
4. taxonomies: ['dei', 'us-gaap'] | us-gaap concepts: 503
5. CY2019Q1I | companies: 3390
   row keys: ['accn', 'cik', 'end', 'entityName', 'loc', 'val']

Quoted as printed from our guide SEC EDGAR API in Python. Apple is only an example filer, not a view on the company.

BUILD / COMPARE / RESULTSMethod first, then results

01 / THE RESEARCH BENCH

First job on the bench: stock research.

All tasks

BUILD/01

Give an agent real data

Pull filings and XBRL financials free from SEC EDGAR: the full Python script, plus its output from a dated run.

Next: read a 10-K with AI

Guide · run on 2026-10-20

SEC EDGAR API in Python

COMPARE/02

Same job, two agents

Grok Bot and Claude Code get the same stock research job and the same inputs, scored on accuracy, cost, speed and setup.

All comparisons

Method published · no results yet

Grok Bot vs Claude Code

RESULTS/03

An AI analyst, graded

Every call is logged before the outcome and graded later, misses included. No hit rate until there is a real record.

Method published · no scores yet

Can AI pick stocks?

Sample research notes, built by AI

Educational content only, not investment advice. We show how we build and test AI research tools; nothing here is a recommendation to buy or sell any security. Do your own research.

02 / HOW WE TEST

Same task. Same inputs. Dated results.

When we compare agents, each one gets the same task and the same inputs, at least three times, and we report how many runs finished, not only the best one. Every result carries its date, screenshots are scrubbed of personal data, and nobody can pay to change a ranking.

Until we have run something ourselves, its page says Not tested yet.

Read how we test

03 / GROK BOT TEMPLATES

The template shelf.

All templates

Grok Bot is one of the personal AI agent platforms we use. Its 245 available public templates are listed here with short summaries we wrote, and every template page says whether we have tested it. So far, none are tested.

Latest guides and comparisons

Rerun: what it is and how to start The self-hostable AI coworkers app (rerun-build/rerun), not the robotics data tool.
Alternatives to Grok Bot Three open-source agents and three hosted ones, side by side, from their own docs.
Claude Code vs Codex Anthropic's coding agent next to OpenAI's Codex, from the official docs: surfaces, models, plans, and how to choose.
OpenMausBot: what it is and how to start The open-source, Grok Bot-style chat app that runs on your own agent CLIs.
Claude Code vs Cowork Who each agent is for, where it runs, and which plans include it, from Anthropic's pages.
Claude Code vs Cursor An editor with Agent, a CLI, and Cloud Agents, next to a terminal coding agent, from the official docs.
Hermes Agent: what it is and how to start What the project's docs say it is, and the commands for a first chat.
OpenClaw: what it is and how to start The Gateway, a first install, the security defaults, and how it differs from Grok Bot.

FOLLOW THE TESTS

Follow the tests

Videos of the runs behind these pages are coming to our YouTube channel. Subscribe to get them first, or follow new pages by RSS.

Personal AI agents.Real work.