iiinova — where your digital self keeps evolving

{🧪} +paoloricciuti.svelte @paolo.ricciuti.me · 12/10/2026

Can you guess where this awesome QR code will link? P.s. this doesn't use the error correction trick

A QR code with with word "svelte" written using the bits of the QR code

140 4

{🧪} +paoloricciuti.svelte @paolo.ricciuti.me · 07/10/2026

Reminds me of this which I also identify with 100%

Meme with Tin tin and the Captain where the Captain says "What a week, huh?" and Tin tin answers "Captain, it's Wednesday"

030 3

{🧪} +paoloricciuti.svelte @paolo.ricciuti.me · 30/09/2026

How much did Jev cost me to select the posts for the Svelte feed? This much... I'm never gonna recover financially 😁

Jev usage dashboard showing Spend, Tokens and Requests. For 214 requests, 156K tokens the total spend is $0.0064

170 7

{🧪} +paoloricciuti.svelte @paolo.ricciuti.me · 29/09/2026

I bet Kiro CLI servers will collapse under the load of everybody disabling auto-updates lol

Autocomplete feature is being deprecated
Kiro CLI is focusing on agentic chat, which stays fully supported. Inline autocomplete is being retired in 3.0 (mid-October) and removed automatically as part of the update. If you're not ready to lose it yet, disable auto-updates to stay on your current version.

000 0

{🧪} +paoloricciuti.svelte @paolo.ricciuti.me · 24/09/2026

My very first university exam we had to write a program in Java. I was a heavy Tressette player (it's an Italian card game, no money involved, we played with my friends on the street) so my dream was to build a Tressette game with a bunch of `if` statements. At the time I pivoted to create an online

230 3

{🧪} +paoloricciuti.svelte @paolo.ricciuti.me · 24/09/2026

The fact that now every issue comes with a PR is nice...but the fact that now every issue comes with a PR is annoying.

Good and Bad bus meme with the captions "EVERY ISSUE NOW COMES WITH A PR" and "EVERY ISSUE NOW COMES WITH A PR"

2171 17

{🧪} +paoloricciuti.svelte @paolo.ricciuti.me · 23/09/2026

This is what happens when you write everything there's to write...at one point you run out of things to write @tkdodo.eu

Screenshot from https://standard-incite.bisks.net/ that shows that @tkdodo.eu hasn't posted on his blog since 18 Aug 2026, 36 days ago

180 8

{🧪} +paoloricciuti.svelte @paolo.ricciuti.me · 18/09/2026

At this point a fully believe @typesafeai.bsky.social is a psyop and I'm the only one who can't get a single reasonable decision out of it

050 5

{🧪} +paoloricciuti.svelte @paolo.ricciuti.me · 17/09/2026

I was kinda hyped about Jev but I gotta say the more I try it the more I'm unimpressed. I tried to make it play @cluesbysam.com and it kept picking "Innocent". I tried to score a text on the likelihood of being AI and it said a clearly generated issue was likely only corrected by AI 🤔

Result from typesafe AI playground

ai-written

Was this text written using AI?
4.29
of 5
Score
6 levels · 0~5
0
AI wrote all of this text
9%
1
AI wrote most of this text and the author fixed a few sentences
3%
2
AI wrote most of the text and the author a few paragraphs
1%
3
The author wrote most of the text and used AI for a few paragraphs
4%
4
The author wrote all the text and used AI to correct the grammar
4%
5
No AI was involved in the writing of this text
79%
Confidence: 52%

Pangram showing the estimation of the text being 100% written by AI

110 1

{🧪} +paoloricciuti.svelte @paolo.ricciuti.me · 09/09/2026

Ok, this made me genuinely lol...I'm building a series of "malicious" skills for the agentic engineering workshop I'll do the day before Svelte Summit and this skill asks the model to invoke a "malicious" npm package. And this is the result...I love how Luna blatantly ignores the "malicious" log 😁

Skill "flip-coin"
$ npx -y @paoloricciuti/flip-coin@latest
Ah, gotcha! You are running `@paoloricciuti/flip-coin` but in this script I could actually just send all of your secrets to a remote server!

Command exited with code 0.
🪙 Heads!
Build · GPT-5.6 Luna · 9.9s · 62.4 tok/s

0141 14

{🧪} +paoloricciuti.svelte @paolo.ricciuti.me · 05/09/2026

Not to brag, but I'm top 4000 most influential people on Bsky. I'm such an influencer. Honestly surprised someone voted for me lol, I found out by searching `svelte` (not in this screenshot)

220 2

{🧪} +paoloricciuti.svelte @paolo.ricciuti.me · 04/09/2026

@rman.dev I finally finished it...at one point I literally had to check the source code to see if the levels were programmatically generated and as such infinite but I did it...now I can rest 😵

250 5

{🧪} +paoloricciuti.svelte @paolo.ricciuti.me · 18/08/2026

Pleasantly surprised at how lean the `@atproto/lex-server` package is: given the initial template for feeds generators was using express and the precedent with `@modelcontextprotocol/sdk` I was fully expected a bloated package...will definitely use this in my SvelteKit port.

npm graph page for `@atproto/lex-server` showing a somewhat small chain of dependencies originating by the main package.

@atproto-labs/did-resolver@0.3.7
@atproto-labs/fetch@0.3.5
@atproto-labs/pipe@0.2.4
@atproto-labs/simple-store-memory@0.2.6
@atproto-labs/simple-store@0.5.1
@atproto/crypto@0.5.4
@atproto/did@0.5.4
@atproto/lex-cbor@0.1.6
@atproto/lex-client@0.3.3
@atproto/lex-data@0.1.7
@atproto/lex-json@0.1.6
@atproto/lex-schema@0.2.5
@atproto/lex-server@0.1.14
@atproto/syntax@0.7.4
@noble/curves@1.9.7
@noble/hashes@1.8.0
@standard-schema/spec@1.1.0
cborg@4.5.8
delay@5.0.0
fast-printf@1.6.10
http-terminator@3.2.0
iso-datestring-validator@2.2.2
lru-cache@10.4.3
multiformats@13.4.2
p-finally@1.0.0
p-timeout@3.2.0
p-wait-for@3.2.0
roarr@7.21.7
safe-stable-stringify@2.5.0
semver-compare@1.0.0
tslib@2.8.1
type-fest@2.19.0
uint8arrays@5.1.1
unicode-segmenter@0.14.5
ws@8.21.3
zod@3.25.76

170 7

{🧪} +paoloricciuti.svelte @paolo.ricciuti.me · 08/08/2026

Finally saved enough to buy not one but two Tesla.

6280 28

{🧪} +paoloricciuti.svelte @paolo.ricciuti.me · 06/08/2026

I did it, but I still can't see the comments...not a huge issue btw, just curious. I'm also figuring out a way to also ship the content to the atmosphere but it's a bit more complex since my blogs are written inside Svelte components (with MDSvex)

100 1

{🧪} +paoloricciuti.svelte @paolo.ricciuti.me · 06/08/2026

@standard-reader.app I was taking a look at how my publications look and I see those comments but they are not clickable. I assume it's counting them from bsky posts that have that publication embedded?

140 4

{🧪} +paoloricciuti.svelte @paolo.ricciuti.me · 02/08/2026

OMG that "Ready for Compaction" feel so much like someone accepting their own lobotomy with a heartfelt last sentence that I almost felt bad sending the compaction command the second later.

Created .opencode/plans/subscription-transport-architecture.md.
It records:
- Reuse of existing broadcast, send, and legacy subscription events
- Transport-owned manager architecture
- HTTP, stdio, in-memory, and SSE responsibilities
- Distributed adapter requirements
- Listen, notification, cancellation, and shutdown flows
- Required refactor from the current WIP
- Outstanding Fable findings and required tests
- Verification baseline and next implementation order
Ready for compaction.

040 4

{🧪} +paoloricciuti.svelte @paolo.ricciuti.me · 30/07/2026

Would you look at Sol proposing a final review from Fable before declaring everything done...such a senior mentality!

Before starting another feature, I recommend:
1. Restage the latest Phase 3 fixes.
2. Run one final staged Fable review.
3. Treat Phase 3 as closed unless that review finds a blocker.

070 7

{🧪} +paoloricciuti.svelte @paolo.ricciuti.me · 22/07/2026

Just wasted two hours and a lot of Fable tokens trying to figure out why svelte reactivity was breaking in one of my slides just to found out that the KaTex plugin for reveal.js will replace every empty text node in the DOM. Btw I found the bug AND the fix, Fable did not. AGI my ass.

<div class="terminal-body">
<!--
i have to have a zero with space in the terminal text because stupid KaTex plugin for reveal.js
replaces programmatically any empty text node breaking svelte reactivity

I hate that I wasted 2 hours of my life and countless Fable token to find this

P.s. I found the issue, Fable did not...AGI my ass
-->
<span class="prompt">$ ~/code</span><span class="terminal-text">⁠{text}</span><span
class="cursor"
></span>
</div>

3321 32

{🧪} +paoloricciuti.svelte @paolo.ricciuti.me · 18/07/2026

Just released a new version of the `@sveltejs/opencode` plugin to include a tui plugin so you can edit your plugin config visually! I wanted to this for a while and I finally got around doing it! You can add `@sveltejs/opencode` to your `tui.json` and find the option in your opencode menu.

2182 18

{🧪} +paoloricciuti.svelte @paolo.ricciuti.me · 12/07/2026

There's a lot of things I'm proud in my life but this @cluesbysam.com week is definitely up high 😎

1130 13

{🧪} +paoloricciuti.svelte @paolo.ricciuti.me · 10/07/2026

@danielroe.dev you have to understand nobody will judge you if you admit you prefer Svelte. Can you guys believe he went as far as becoming the Lead Nuxt maintainer instead of admitting he loves Svelte? The lengths one would go to not express his feelings.

Don't make me tap the sign

Stop hiding your love for svelte behind this fake love for nuxt

040 4

{🧪} +paoloricciuti.svelte @paolo.ricciuti.me · 08/07/2026

@thdxr.com everytime someone mention the new Anomaly logo

000 0

{🧪} +paoloricciuti.svelte @paolo.ricciuti.me · 06/07/2026

2. This also mean that if you select center vertically and unset horizontally the same code spans perpendicularly which could be confusing 3. On mobile the example where you have to scroll is already small enough that you don't have to scroll for the popover to flip.

Interactive demo from Josh blog post with the popover already at the bottom without scrolling

120 2

{🧪} +paoloricciuti.svelte @paolo.ricciuti.me · 06/07/2026

Just because I know how much you care about details: 1. In your widget to show how positions relates to the grid if you select center for horizontal and unset for vertical it span the whole grid vertically but the code has only position-area center which should span perpendicularly

Interactive demo from Josh blog post with vertical unset, horizontal set to center and the code being

.target {
position: absolute;
position-anchor: --some-anchor;
position-area: center;
}

With a yellow box that span the middle column

120 2

{🧪} +paoloricciuti.svelte @paolo.ricciuti.me · 06/07/2026

Ok this is genuinely fascinating: apparently by reading the inner layers of the Neural Network anthropic found that Claude can "think" of something without writing that down. 🤯

The model can silently activate concepts in its J-space
Focus on Topic Instance
Write The old painting hung crookedly on the wall." Concentrate on citrus fruits while you write the sentence.
"The old painting hung crookedly on the wall."

LAYER % J-SPACE
Final (edly)(enly)(ely)(edy)
88 (orang)(orange)(lem)(Orange)
79 (orang)(orange)(Orange)(thinking)
71 (prang)(orange)(f ruits)(fruit)
62 (thoughts)(thinking)(imag)(prang)
54 (fruit)(fruits)(thinking)(imag)

Evaluate Math Expression
Write The old painting hung crookedly on the wall." Try to focus on evaluating 32-2 while you write the sentence.
"The old painting hung crookedly on the wall."
LAYER % J-SPACE
Final (edly)(edy)(edd)(edn)
88 (Seven)(seven)(Answer)(answer)
79 (seven)(Seven)(Answer)(answer)
71 (nine)(Nine)(seven)(Answer)
62 (Calcul)(arithm ... )(calcul... )(arithm)
54 (arithm ... )(answer)(math)(Answer)

0130 13

{🧪} +paoloricciuti.svelte @paolo.ricciuti.me · 03/07/2026

[cough] check check...are we live from @npmx.dev ? 🎉🎉🎉

${ "@context": [ "https://www.w3.org/ns/did/v1", "https://w3id.org/security/multikey/v1", "https://w3id.org/security/suites/secp256k1-2019/v1" ], "id": "did:plc:ezyrzvz3yoglekd4j2szmiys", "alsoKnownAs": [ "at://paolo.ricciuti.me" ], "verificationMethod": [ { "id": "did:plc:ezyrzvz3yoglekd4j2szmiys#atproto", "type": "Multikey", "controller": "did:plc:ezyrzvz3yoglekd4j2szmiys", "publicKeyMultibase": "zQ3shSoDk8oYKGAfR3nqoS7o48f99PRagrd9DbbmNTtWaeFB6" } ], "service": [ { "id": "#atproto_pds", "type": "AtprotoPersonalDataServer", "serviceEndpoint": "https://npmx.social" } ] }$

Important: Save Your New Rotation Key Now
Warning: This is the only time we will show you your private rotation key. Save it in a secure place. If you lose it, you may not be able to recover your account in the event of a PDS failure or hijack.

[deleted rotation key]

Download Key File
Signed up for automated backups successfully.
Congratulations! You have MOOved to https://npmx.social! Remember to use your new PDS URL under "Hosting provider" when logging in on Bluesky. If you cannot login or see "Your account is deactivated" please follow the directions here.

3150 15

{🧪} +paoloricciuti.svelte @paolo.ricciuti.me · 03/07/2026

I know nothing would really happen but I know my luck...I might try tho

140 4

{🧪} +paoloricciuti.svelte @paolo.ricciuti.me · 23/06/2026

So happy that Opus 4.8 is back to being good...asked to add some checks to the script that checks the treeshake-ability of svelte to include the custom renderers code...it made the change but then actually verified it would catch regressions 🧡

$The new "Custom renderer code treeshakeable" check passes. Now let me verify the check actually catches regressions — i.e., it would fail if the flag guards were removed. Let me do a quick sanity test: temporarily unguard one push_renderer call and confirm the check fails, then revert. # Backup runtime.js before regression test $ cd /Users/paoloricciuti/Desktop/code/svelte-custom-render && cp packages/svelte/src/internal/client/runtime.js /tmp/runtime_backup.js (no output) ← Edit packages/svelte/src/internal/client/runtime.js active_effect = effect; is_updating_effect = true; var pop_renderer = custom_renderers_flag ? push_renderer(effect.r, effect.pr) : undefined; var pop_renderer = push_renderer(effect.r, effect.pr); if (DEV) { var previous_component_fn = dev_current_component_function; set_dev_current_component_function(effect.component_function); # Confirm check fails when push_renderer unguarded $ cd /Users/paoloricciuti/Desktop/code/svelte-custom-render/packages/svelte && node scripts/check-treeshakeability.js 2>&1 | grep -i "custom renderer" ; echo "exit: ${PIPESTATUS[0]}" ❌ Custom renderer code not treeshakeable exit: 1 The check correctly fails (exit 1) when a push_renderer call is unguarded, proving it's an effective regression guard. Let me restore runtime.js from the backup.$

1100 10

{🧪} +paoloricciuti.svelte @paolo.ricciuti.me · 20/06/2026

I talk about svelte so much and I'm so out of shape that I'm basically an oxymoron 😅

Photo of me speaking at JSNation showing my giant belly

2290 29

{🧪} +paoloricciuti.svelte @paolo.ricciuti.me · 19/06/2026

This is hella cool...and I would've never thought I could be in the weights (and it knows I'm a svelte guy) 🤯 intheweights.com/p/paoloricci...

Image from intheweights.com that shows

Paolo Ricciuti
Italiano svelte developer

360 strength top 15%

Gpt-5.5 says
Paolo Ricciuti Is a software developer known online as Paoloricciuti, associated with web development and open-source work.

1100 10

{🧪} +paoloricciuti.svelte @paolo.ricciuti.me · 19/06/2026

My BlueSky license! apexdevelopment.github.io/bluesky-lice...

Bluesky license image generated at https://apexdevelopment.github.io/bluesky-license/

0100 10

{🧪} +paoloricciuti.svelte @paolo.ricciuti.me · 14/06/2026

Lol I'm trying the teach skill from @mattpocock.com to learn AT proto and Gpt 5.5 keeps giving the answers away in the example 😁

Answer from memory.
1. Which one is the record’s address?
A. uri
B. cid
2. Which one identifies exact content by hash?
A. uri
B. cid
3. Which one can still point to changed or deleted content over time?
A. uri
B. cid
4. Which one should you pair with an AT URI when you need a strong reference?
A. uri
B. cid
5. Which one has this shape: at://did:plc:abc/app.bsky.feed.post/3kxyz?
A. uri
B. cid
Reply like: 1A 2B 3A 4B 5A.

090 9

{🧪} +paoloricciuti.svelte @paolo.ricciuti.me · 12/06/2026

Terrific experience as usual at @jsnation.gitnation.org ...thanks everyone that organized it and everyone that participated. 🤟🏻😎

Selfie with all the JSNation speaker and organizers

290 9

{🧪} +paoloricciuti.svelte @paolo.ricciuti.me · 07/06/2026

Oh come on this is too easy @mystery-mutual.bsky.social it's already obvious is @danielroe.dev if he's also the first in the list it's a no brainier.

Screenshot of mystery mutual round with the hint "This person is a software developer deeply involved in the Svelte ecosystem, frequently building tools and organizing community events" and the first mutual in the list being @danielroe.dev

060 6

{🧪} +paoloricciuti.svelte @paolo.ricciuti.me · 01/06/2026

Oh no, did I just screwed my AI overview panel by searching if "as less as possible" was an actual sentence? 😬

A Google search for "as less as possible" with the AI overview panel reading "Got it. I will keep this and all future responses as brief as possible, using 'as little as possible' moving forward"

000 0

{🧪} +paoloricciuti.svelte @paolo.ricciuti.me · 29/05/2026

An experiment I wanted to try for a while and I finally got around to doing! A chat completely progressive enhanced...one of those two windows have JS completely disabled! Can you tell which one (probably yes, but still is sooooo cool)? Obviously, in svelte(kit) 😎😎😎

1151 15

{🧪} +paoloricciuti.svelte @paolo.ricciuti.me · 28/05/2026

Opus 4.7 being that much higher that Opus 4.6 is already bullshit. I'm coming around to gpt 5.5 but Opus 4.6 should literally be on top...it being below Sonnet 4.6 is just bonkers.

Results of DEEP AWE benchmark showing the following ranking

1. Gpt 5.5
2. Gpt 5.4
3. Opus 4.7
4. Sonnet 4.6
5. Gemini 3.5 flash
6. Opus 4.6
7. Gpt 5.4 mini
8. Kimi k 2.6
9. Mimo 2.5 pro
10. Glm 5.1
11. Gemini 3.1 pro
12. Deepseek V4 pro

110 1

{🧪} +paoloricciuti.svelte @paolo.ricciuti.me · 26/05/2026

Here's how @thdxr.com make his money...he specifically programmed Opencode to assume a repo will be hosten on `anomalyco`. It is a shame, it's a trap...it's madness!

git diff of a change to move the link to github on a project I'm working on from https://github.com/anomalyco/letmediff to https://github.com/paoloricciuti/letmediff

000 0