Evan Batky

Making videos about AI 🤖

Helping businesses save time and money 💴

[Evan Batky @evan_batky]

Can Claude Code Run While You're Away?

Claude Code Remote Control puts the session on your phone while it still runs on your home computer, so your files, skills, and MCPs stay available.

#ClaudeCode #RemoteControl

[Evan Batky @evan_batky]

Claude Code Flags Underperforming Meta Ads Daily

Every morning, this Claude Code mod pulls the previous day's Meta Ads results, compares them with its own average, and writes a brief note in the panel. It never changes the budget.

#ClaudeCode #MetaAds

[Evan Batky @evan_batky]

@HermesWatcher Having a separate signed-in session is the part that matters to me. My agents get way more reliable once they stop fighting me for the same browser tab

[Evan Batky @evan_batky]

@BitmonkCrypto Same here. What worked for me was keeping a running list of those new ideas and only building the ones I'd otherwise do by hand every week. Most of the rest can wait

[Evan Batky @evan_batky]

Turn Shopify's Default Theme Into Your Own Brand

A basic Shopify template can become a branded storefront with a stronger homepage and clearer collections. Claude handles the changes while you approve and test each step.

#Shopify #AIWebDesign

[Evan Batky @evan_batky]

@DestinyEss1241 I've found splitting agents by role like this makes debugging so much easier. The tricky part is the handoff between them, so I keep a shared log every agent writes to.

[Evan Batky @evan_batky]

@beecrypto_68 On-chain identity is interesting, but I've found memory is the hard part with agents. How does it decide what's worth keeping versus what just piles up ?

[Evan Batky @evan_batky]

@TomTanzim Running out of credits mid-workflow is a sneaky failure mode. What worked for me is adding a balance check and an alert node before the expensive steps run

[Evan Batky @evan_batky]

@fatma_zaidi1 For me it's productive when the agent runs a long task and I just approve steps from my phone. Actually writing prompts on a phone keyboard is another story

[Evan Batky @evan_batky]

@mr_r0b0t I've found most agent browsing doesn't need full rendering, just the DOM after JS runs. Curious how it handles sites that fingerprint headless browsers

[Evan Batky @evan_batky]

@darebude Claude Code for anything real, Cursor when I want to see every diff. The prototyping tools are great for demos, but I always end up rebuilding in a proper repo

[Evan Batky @evan_batky]

@usepushary @_HimanshuBuilds 🙌

[Evan Batky @evan_batky]

@LoadoutGuide Amazing.

[Evan Batky @evan_batky]

4 Claude Code Features, 4 Different Jobs

Skills load instructions for a task. Hooks run at set points. Subagents handle side work in a separate context. Mods change Claude Code itself.

Which one do you use most?

#ClaudeCode #AIagents

[Evan Batky @evan_batky]

@buildwbhoomika Simple AI-based receipt scanner

[Evan Batky @evan_batky]

@Eminweb3 Running out of API credits mid-run is a sneaky failure. I've started adding balance checks and alerts at the start of workflows so nothing dies silently

[Evan Batky @evan_batky]

@ruposhamin What I've found is that measurable results start with picking one painful workflow and tracking time saved before adding agents or RAG on top

[Evan Batky @evan_batky]

@skyprince70 @BAI_AGI @justinsuntron GUI is a great on-ramp. What I've found is the hard part for non-devs isn't setup, it's knowing what to do when a step fails, so clear error handling matters a lot

[Evan Batky @evan_batky]

@ShaneBOly Same rule here. I also have it write a one-line summary per file changed, so vague answers like that get caught before I commit anything

[Evan Batky @evan_batky]

@DaneBuilds Same pattern here. The narrow tools that survive will be the ones holding data or workflows people can't easily rebuild, not just a nice UI over one feature

[Evan Batky @evan_batky]

@JamesMalsawm Feels noticeably better at staying on track over long sessions for me. Planning first, then letting it build in small chunks, has been the sweet spot so far