dvsj

@GoIvanti recently suffered a series of 0-day vulnerabilities. One of those - high severity authentication bypass. But we can all learn from them, so here's a quick recap. 🧵x2 Takeaway #1: Use 𝚎𝚚𝚞𝚊𝚕𝚜 for string comparison. Not 𝚜𝚝𝚊𝚛𝚝𝚜𝚆𝚒𝚝𝚑 or 𝚌𝚘𝚗𝚝𝚊𝚒𝚗𝚜.

@justinmk This is useful in plugins etc that have better context to handle user actions. Con: plugins need a shared vocabulary (to understand what "autocorrect accepted" means) Tldr @justinmk implemented atoms (structured inputs) and multiline is just a feature to show it off (I think)

@justinmk But if we had atoms (structured input), we could have avoided clobbering "let getRekt| there |be" Cursor1: autocomplete accept worked Cursor2: show error: "no autocomplete found to accept, cancelling"

Thanks @justinmk I didn't get this, but after some reading I understand in retrospect 2 ways an editor can record actions: - record exact keys pressed ("pressed g, e, t, enter") - record atomic actions w intent ("typed get, accepted first autocomplete suggestion with enter")

@amorriscode haha, looking forward to checking it out!

@dvsj_in yeah my domain expired this morning🤦

who’s next?

@gm_mertd All fun and games till they start shifting layouts on more complex UIs

Stop using alerts and use in-place transformations

@pantharshit007 @CloudflareDev idk, the purpose is to slow you down for a few seconds if it doesn't do that, what's the point of the input box?

This is what building while keeping UX in mind means 🫡 @CloudflareDev

@gopikl "Good" is doing a lot of the heavy lifting here :P

To some extent, I still believe - "build it and they'll come". There are so many folks who say distribution is the moat, but there are enough examples to prove product drives distribution when it's good.

@stevekrouse Sending you a DM, save it for later :P

there's an effervescent feeling when there's an agent working for me that's similar to the feeling of an unread message from someone you admire - a little treat for me to check in on later

@zeeg is this sentry swag?

Good night little buddy

@kousikmahadeva1 The zoho effect? :P Thanks bro 🙌🏾😁

@dvsj_in That's one whole H1 delayed! 😂 Nice write up!

Better two quarters late than never? My 2025 - year in review 😬 dvsj.in/year-in-the-re…

@PrajwalTomar_ Nope. Unintuitive, but using Fable doesn't guarantee skills will be better than the same thing done with Opus. Fable's strength is orchestration. Opus still beats it sometimes at specific tasks. x.com/dvsj_in/status…

Fable 5 goes pay per use in 2 days. Make it train its replacement before it does. Have Fable 5 write Skills for the models you'll actually keep using, so its judgment stays even after the model goes behind credits. The best version from the thread: tell Fable it's "a retiring principal engineer" leaving a complete skill library behind for the team. → It audits your repo like an incoming senior engineer → Writes 10-16 SKILL files: debugging playbook, change rules, the failure stories that cost real time → Opus and Sonnet then run that playbook on cheaper sessions, at Fable's standard Fair warning from the thread: the run can eat 30% of your weekly Fable usage. Still worth it. Full prompt in comment 1, my Fable 5 playbook below. Bookmark this.

1. Improving skills is valuable. 2. Fable DOES NOT guarantee better skill improvement than Opus. 3. Even the best skills miss issues (rank #1 missed a sev0) I HATE this, my mental model would've been so much better if Fable just won. You won, @zeeg. We still need to read code.😒

1. Everything being same, model Fable (14) beats Opus (9) 2. Model being same, skill enhanced by Opus (16) beats Fable (13) 😶 3. Better model makes up for bad skills? Fable (14) vs Opus (9) on same trash skill (trash skill bc it ended up being half the size after enhancement)

I think we all agree better model == better output? Not a stretch to think better model writes better skills But I'm gonna test this