←
Bart R. McDonough9,627 posts
Bart R. McDonough avatar

Bart R. McDonough

@BartMcDonough

Father to Russell, Former CEO of Agio (exited), Authored two books, Oklahoma native, OU supporter, Yale MBA, aspiring golfer, living on the coasts and mountains

◷ Joined January 2008▧ 1,440 Photos and videos
Bart R. McDonough avatar
Bart R. McDonough@BartMcDonough·Nov 17

The Optimizely hack is serious. Malicious code was distributed through major websites you probably visit. Two things I recommend for iPhone users: 1. Restart your iPhone. Rebooting can clear malware running in memory, although it isn’t a guaranteed fix. 2. Update your iPhone. Go to Settings → General → Software Update and install the latest iOS version. This attack affected websites used by major brands, including CNN, BBC, Zoom and Salesforce. Take a minute and do both.

Ebi avatar
⟳ Bart R. McDonough retweeted
Ebi@EbiThinks·Oct 30

True.

0:10
Bart R. McDonough avatar
Bart R. McDonough@BartMcDonough·Jul 21

65% of product-engineering PRs coming through Claude in Slack is a serious number. Now I want the denominator: accepted without edits, reviewer time, reverts, and production defects. PR share can go up while the engineering bottleneck moves into review.

We talked about Claude Code, Claude Tag, Fable, coding agent security, evals, tool design, and how Anthropic use these tools themselves Claude Tag (Claude Code via Slack) is already landing 65% of the product engineering PRs for the Claude Code team!

Bart R. McDonough avatar
Bart R. McDonough@BartMcDonough·Jul 20

Secret scanning found 20,000 alerts. GitHub still needed nine months to validate credentials, find owners and fix the systems. Detection was the fast part. Ownership was the work.

GitHub@githubJul 20

20,000+ secret scanning alerts across 15,000 repos. Nine months later, GitHub reached inbox zero.

Bart R. McDonough avatar
Bart R. McDonough@BartMcDonough·Jul 15

Agents are creating work faster than teams can review it. Track intervention rate, review time, rollback rate and cost per accepted change. Lines of code will lie to you.

Bart R. McDonough avatar
Bart R. McDonough@BartMcDonough·Jul 15

Production usage is more useful than another benchmark. Requests, tokens and spend tell me where a model actually belongs in a workflow.

Vercel@vercelJul 14

AI Gateway leaderboard data is now open. Real production usage across models, labs, apps, and providers, updated daily. Download, query, and cite it under CC BY 4.0. vercel.com/changelog/open-da…

Bart R. McDonough avatar
Bart R. McDonough@BartMcDonough·Jul 14

I want the receipt from every agent run. What changed. What failed. What it cost. Which permissions it used. What still needs a person. A green check is not enough.

Bart R. McDonough avatar
Bart R. McDonough@BartMcDonough·Jul 14

Agents expose application debt very quickly. Missing APIs, undocumented state and unclear ownership show up before the automation does.

Amazon Web Services @awscloud Jul 13 Article When the bottleneck is not your agent, but the applications it depends on Introduction You design an agentic workflow to automate order fulfillment without first evaluating whether your applications are ready for agent interaction. The agent monitors incoming orders,

Bart R. McDonough avatar
Bart R. McDonough@BartMcDonough·Jul 13

Agent memory without provenance is a security problem. Show me what created the memory, when it changed, and where it is being used. Let me delete it. One bad email should not become trusted state.

Bart R. McDonough avatar
Bart R. McDonough@BartMcDonough·Jul 13

ChatGPT Sites will create useful internal tools very quickly. It will create orphaned ones just as fast. Someone still has to own the data, changes, support and cleanup.

jason@jxnlcoJul 13

ChatGPT Sites is now in public beta. You can turn a prompt, file, or rough idea into a dashboard, project tracker, report, prototype, or lightweight app: build and edit directly in ChatGPT Work or Codex test in a private preview publish and share with a URL Rolling out across paid plans. Enterprise admins can control public publishing.

Bart R. McDonough avatar
Bart R. McDonough@BartMcDonough·Jul 12

GPT-5.6 is excellent. The Codex usage math is not. I’m hitting five-hour limits in 60–90 minutes doing work that fit under 5.5. Better model. Worse daily workflow.

Bart R. McDonough avatar
Bart R. McDonough@BartMcDonough·Jul 11

The model is not the product anymore. For serious AI operators, the product is the work loop: goal context tools state sandbox approval receipt cost That is what decides whether agents become daily workflow.

Bart R. McDonough avatar
Bart R. McDonough@BartMcDonough·Jul 11

ChatGPT Work is the important line because it moves AI from answer box to work surface. The operator question changes: Which tools can it touch? What state does it keep? Where does it ask for approval? What receipt comes back?

Introducing ChatGPT Work, a new agent in ChatGPT powered by Codex and GPT-5.6. It can take action across your apps and files, stay with a project for hours if needed, and turn a goal into finished work. It’s a whole new way to get work done.

1:40