Via How I AI

“I’ve built a skill called Improve, which is basically a very critical auditor for these… I ask: is this real? Is this powerful? How does this adapt to what we have now? Is this needed? And then it helps me review it, and together we basically decide whether we want to build it or not.”

— Daniel Blum | I built a Claude Cowork system that does a week of PM work in a day | How I AI, watch at 30:35

---

Keeping up with AI now means wading through an endless flood of advice. Posts on X and LinkedIn, newsletters, and blogs all insist that you must have this setup and build that skill. Daniel Blum, a product manager at Melio, found it overwhelming. He didn’t want to drown in it, but he didn’t want to ignore it either.

His solution is a skill he calls Improve. He sends anything interesting that crosses his feed, usually from his phone into a Slack channel that Claude checks. Claude evaluates each item against four questions: is it real, is it powerful, does it fit what he already runs, and does he need it?

The filter is built around fit, not quality. A tip can be accurate but not useful for your setup. Two of the four questions test the idea against your specific situation and context.

Daniel ran the Improve skill on another episode of How I AI that talks about designing loops. Claude called it a good video, then noted that loops are essentially scheduled tasks, which Daniel already has. However, Claude highlighted that Daniel doesn’t have the goal loops the episode describes. The audit surfaced a specific gap in Daniel’s setup, which is more useful than a general “this is a great tip” verdict.

What makes this work:

• Capture is cheap. Sharing a link from your phone takes seconds, so nothing relies on remembering it later.
• Judgment stays shared. Claude researches and critiques, and Daniel makes the call. They decide together whether to build, adapt, or skip.
• The output is a decision. A summary tells you what a tip says. An audit tells you whether to act on it.

The auditor’s default is to say no, so a tip has to prove its worth before it is adopted. Without a filter, every new tip competes for your attention on equal terms, and the loudest wins. With an auditor, the cost of ignoring hype drops to nearly zero, and you get to focus on the few ideas that fill a real gap.

Daniel calls Improve “critical today … to keep up and become better”. The influx of tips won’t slow down, so the answer is a filter that only keeps what’s worth your while.

---

Things to Try Today

Set up your intake:

• Create one place where AI tips go, such as a Slack channel, a notes file, a Telegram chat, or a chat with Claude.
• Send it the next three tips that catch your eye before acting on them.

Write your audit questions:

• Paste Daniel’s four questions into a prompt: Is this real? Is it powerful? How does it fit what I have now? Is it needed?
• Add one of your own, such as “What would I have to stop doing to make room for this?”
• Ask Claude to answer each question about your tip and finish with one verdict: adopt, adapt, or ignore.

Run it once, then decide:

• Audit the last tip you were tempted to follow.
• Give it 10 minutes. If the verdict is “adopt” or “adapt”, schedule the work. If it’s “ignore”, let it go.