Blog
Practical notes from Digital Meld on AI, automation, cloud, data, and operating systems that have to work in the real world.
Latest Articles
Sep 9, 2026 | Brad Groux
From GPT-5.6 Sol to GPT-6 Astra: Why Am I Repeating Myself?
Six days of Astra records, 78 correction episodes, and the instructions I kept having to repeat after moving from my GPT-5.6 workflow.
Sep 8, 2026 | Brad Groux
GPT-5.6 Sol vs. GPT-6 Astra: A Second Audit of Veritas Kanban
I audited Veritas Kanban with GPT-5.6 Sol and then GPT-6 Astra, both at medium reasoning. Here is what the sequential passes found and changed.
Sep 3, 2026 | Digital Meld
AI Still Needs Your Judgment, with Doug Gabbard | Episode #0037
Doug Gabbard joins Brad and Robert Groux to explore practical AI workflows, SharePoint agents, photography, and why human judgment still matters in daily work.
Aug 28, 2026 | Digital Meld
Inside OpenClaw: GitHub Talks With the Maintainers Building and Securing It
Brad Groux joined GitHub, OpenClaw creator Peter Steinberger, and fellow maintainers to discuss the work behind OpenClaw's rapid growth: governance, security, model neutrality, community, and building in public.
Aug 21, 2026 | Brad Groux
Questions Before Platforms: Building the Data Foundation AI Actually Needs | Episode #0036 by Start Small, Think Big
AI readiness starts with better questions, shared business definitions, current knowledge, and lightweight governance—not another platform. Brad Groux and Sid Atkinson explain how leaders can build that foundation without slowing the business down.
Aug 18, 2026 | Digital Meld
Who Is Paying for the AI Boom?
AI is already useful. The financing behind frontier labs is increasingly circular, provider revenue remains hard to isolate, and public evidence does not support planning a business around near-term AGI or ASI.
Jul 29, 2026 | Digital Meld
Using Personal AI in 2026: From Business Context to Working Software With OpenClaw and Codex
A practical workshop walkthrough showing how durable business context, a challenged PRD, scoped issues, a bounded goal, human review, and verification turn OpenClaw and Codex into a repeatable software delivery workflow.
Jul 21, 2026 | Digital Meld
How We Use QuickBooks, Agentic AI, and Automation to Clean Up Accounts Payable and Bookkeeping
QuickBooks can be the financial source of truth, but much of the work happens before and after a transaction reaches the ledger. Here is how Digital Meld uses agentic AI, deterministic checks, and human approval to reduce manual AP and bookkeeping work without handing financial authority to a model.
Jul 17, 2026 | Digital Meld
Business AI Starts With the People Who Know the Work
A practical lesson from GitHub Rubber Duck Thursday: useful business AI starts with the people who know the work, then turns that knowledge into process, checks, and reviewable artifacts.
Jul 15, 2026 | Digital Meld
The Incident Plan Has to Exist Before the Incident
A repository attack workshop is a useful reminder for open source maintainers and businesses alike: incident response plans only work when roles, evidence, credentials, communications, and closure criteria are ready before the incident starts.
Jul 8, 2026 | Digital Meld
KARS Makes Agent Infrastructure a Microsoft Stack Conversation
KARS gives Microsoft-focused teams a practical reference for running agents with real runtime boundaries: per-agent sandboxes, governed egress, router-held credentials, audit, policy, and Kubernetes-native operations.
Jul 6, 2026 | Brad Groux
AI Adoption Is a Partner Journey, Not a Turnkey Product
Microsoft Frontier Company is another signal that business AI adoption is moving from model access to hands-on implementation. Useful AI has to be fitted around a business, its people, and its workflows.