Yuval Dolev

Programmer, Foodie.

[Yuval Dolev @_ydolev_]

What happened to Dispatch and Computer Use in Claude Team plans?
They used to be available but aren't anymore now...

[Yuval Dolev @_ydolev_]

Notion exposed as a file system for optimized agent interaction

[Yuval Dolev @_ydolev_]

Looks like GPT 5.6 Sol is so good that Anthropic had no choice but to keep Fable in the subscription plans

[Yuval Dolev @_ydolev_]

Used GPT 5.6 Sol with Sites yesterday
Finally - GPT is great at frontend as well
Amazing results!

[Yuval Dolev @_ydolev_]

Claude's weekly limits got quietly nerfed. Yesterday my reset was 2 days out. Today it's 4. The timer is running backwards.

[Yuval Dolev @_ydolev_]

@Jason I’m in

[Yuval Dolev @_ydolev_]

Here’s the wild part:

I subscribed to ChatGPT Pro normally, then checked the subscription settings, opened the cancellation flow just to see the options, and got offered the next month free.

So it ended up being:
Month 1 = paid
Month 2 = free

Definitely did not expect that 😂

[Yuval Dolev @_ydolev_]

I accidentally found a way to get 2 months of @ChatGPTapp Pro for the price of 1.

No coupon.
No referral.
No sketchy workaround.

Just something unexpected inside the subscription flow.

How it happened and how you can do it too👇

[Boaz @boazbe]

אני דיי בטוח שמי שמתחיל פרויקט תוכנה חדש היום ולא משתמש בRust הולך להפסיד. לכתוב תוכנה לא עולה בזמן יותר, אז אין הצדקה להסתפק בביצועים גרועים יותר שנותנות שפות אחרות.

[ThePrimeagen @ThePrimeagen]

which company's models do you use on the daily?

[Yuval Dolev @_ydolev_]

@leojrr https://t.co/dEv1Virfrs

[Yuval Dolev @_ydolev_]

I’m building an app in the travel space and recently used GitHub Copilot Agent + Claude Opus to bootstrap a dataset (~500 places in Copenhagen: cafes, restaurants, sights, etc).

What I found while testing was… unexpected 🧵

[Yuval Dolev @_ydolev_]

@valigo https://t.co/dEv1Virfrs

[Yuval Dolev @_ydolev_]

I’m building an app in the travel space and recently used GitHub Copilot Agent + Claude Opus to bootstrap a dataset (~500 places in Copenhagen: cafes, restaurants, sights, etc).

What I found while testing was… unexpected 🧵

[Yuval Dolev @_ydolev_]

@theo https://t.co/dEv1Virfrs

[Yuval Dolev @_ydolev_]

I’m building an app in the travel space and recently used GitHub Copilot Agent + Claude Opus to bootstrap a dataset (~500 places in Copenhagen: cafes, restaurants, sights, etc).

What I found while testing was… unexpected 🧵

[Yuval Dolev @_ydolev_]

@Param_eth https://t.co/6EKeBsDCh1

[Yuval Dolev @_ydolev_]

I’m building an app in the travel space and recently used GitHub Copilot Agent + Claude Opus to bootstrap a dataset (~500 places in Copenhagen: cafes, restaurants, sights, etc).

What I found while testing was… unexpected 🧵

[Yuval Dolev @_ydolev_]

Takeaway for anyone building agentic systems:

If reality matters:

✅ Require external sources
✅ Use geocoding APIs
✅ Demand citations
✅ Validate coordinates
✅ Add sanity checks

Never trust generated location data by default.

[Yuval Dolev @_ydolev_]

I’m building an app in the travel space and recently used GitHub Copilot Agent + Claude Opus to bootstrap a dataset (~500 places in Copenhagen: cafes, restaurants, sights, etc).

What I found while testing was… unexpected 🧵

[Yuval Dolev @_ydolev_]

My working theory:

The agent interpreted this as data synthesis, not data retrieval.

Instead of “get real Copenhagen places,” it effectively did:
“generate objects that resemble Copenhagen places.”

Small semantic gap. Huge consequences.