Fabio Biondi / GDE - MVP

Name: Fabio Biondi / GDE - MVP
Address: Trieste, IT
Telephone: +393381196284

Home
Italia
Trieste, Friuli-Venezia Giulia
Fabio Biondi / GDE - MVP

Chiama Indicazioni Sito Web Contatta l'azienda

Google Developer Expert (GDE) | Microsoft MVP | React & Angular Instructor

I'm an active contributor in the italian front-end communities with 10+ years of experience in developing enterprise level applications and components. Especially focused in front-end applications and all related aspects, both visuals and architecturals, I daily use cutting-edge libraries and frameworks offering training, development, code review and consulting on Angular, React and several other technologies. I also collaborate with many training italian companies in order to organize events, training courses and campus.

13/10/2026

Once again this year, I'll help to select the speakers for Web Day 2027!

- Milano - 25 march 2027
- 3 parallel tracks: 18+ speakers
- Tickets currently at JUST €25 (there is the promo: €50 for 2 people!)
- Previous editions: 250/300 attendees

The Call for Papers is officially open and promo tickets are up for grabs!

Event and Tickets: https://www.webdayconf.it/

CFP: https://sessionize.com/web-day-2027/

12/10/2026

This Saturday, I had the privilege of taking the stage at Google DevFest Milano to deliver a masterclass on Generative UI to a small group of attendees.

We explored how to craft dynamic, runtime UI directly from prompts or audio inputs using Gemini, diving deep into three distinct approaches:

- Building from scratch with Gemini SDK using tools / function calling
- Leveraging an Angular Gen UI framework: Hashbrown
- Architecting MCP apps and MCP UI

A massive shoutout to the organizers: DevFest Milan was an absolute amazing in event curation. The energy was electric and truly global, bringing together incredible speakers from Germany, Australia, New York, UK, and so on.

On a fun note, Google Italy hosted an unforgettable pizza-making masterclass for us the day before (see photos : )
Such a brilliant, human touch to kick off the weekend.

A thank you to the entire GDG Milano team and Google Italy for the warm welcome and this incredible opportunity.

05/10/2026

Generare una UI a runtime con un semplice prompt o un comando vocale: cosa mai potrà andare storto? 😅
Non resta che sperare, o pregare, che tutto funzioni. E quale location migliore di quella che vedete in foto per parlarne?

Battute a parte, ho avuto l'onore di partecipare per la prima volta come speaker alla Dev Fest Modena.
Nel mio talk sulla UI Generativa ho illustrato i principi di base e mostrato, tra codice e demo, due approcci agli antipodi:
1) Creazione dell'interfaccia interamente lato client con Gemini e Function Calling.
2) Gestione lato server tramite MCP Apps

Ad ogni modo, la Dev Fest è stata pazzesca, in una cornice fantastica nel pieno centro di Modena. Lo staff e gli organizzatori del GDG Modena sono stati impeccabili: una macchina perfetta. Tanti amici, decine di speaker e contenuti di altissimo livello.

Un ringraziamento di cuore a Maurizio, Mirko, Davide e a tutti gli amici del GDG Modena (perdonatemi se non vi nomino tutti) per questa bellissima esperienza e per aver accettato la mia proposta di talk all'ultimo momento.

PS: Eventi del genere non servono solo a imparare e condividere. Di giorno ci si confronta sui temi di lavoro, ma il post-evento è altrettanto fondamentale: è lì che i contatti si trasformano in relazioni autentiche, tra cene, risate, situazioni imbarazzanti e contesti al limite dell'assurdo. Dico bene Greg, Ludovico (& friend di cui non ricordo il nome 🤣)?

29/09/2026

Ever had to use a POST request just because your search filters were too complex to fit cleanly into a GET query string?

Angular is set to solve this dilemma with `HttpClient.query()`, bringing support for the new HTTP QUERY method (RFC 10008) directly into the framework (PR #70858).

Here is why this is such a welcome addition:

• GET is safe and cacheable, but serializing deeply nested objects into query parameters is clunky and hits URL length limits.
• POST allows rich JSON request bodies, but implies mutation: meaning no default caching and no safe automatic retries.
• QUERY gives you the best of both worlds: safe and idempotent semantics like GET, paired with a structured request body like POST.

The syntax fits right into the existing HttpClient API:

private http = inject(HttpClient);

search(filters: ProductFilters) {
return this.http.query('/items', {
category: 'books',
price: { min: 10, max: 50 },
tags: ['angular', 'ai'],
sort: [{ field: 'rating', dir: 'desc' }]
});
}

It is also designed to work seamlessly across the Angular ecosystem:
• SSR TransferCache: Included by default, avoiding duplicate fetches during client hydration.
• XSRF Protection: Recognized as non-mutating, skipping redundant anti-forgery token overhead.
• Standard Overloads: Supports existing options like `observe` and `responseType`.

The catch? Since it is a new HTTP method, your backend, CDNs, API gateways, and WAFs must also support RFC 10008 before you can roll it out to production.

Are you looking forward to finally replacing search POST endpoints with QUERY?

28/09/2026

Opus 5.5 brings a massive shift in how we work with AI across Claude Code and the Claude app.

We’re finally moving away from micromanaging every single prompt and babysitting the output. Instead, the dynamic has shifted to steering autonomous workflows: defining clear boundaries, setting clear success criteria, and letting the model do the heavy lifting.

Inspired by Addy Osmani’s recent breakdown, here are some practical, game-changing habits to adopt right now:

👉 1. Define what “done” looks like, then step back
Instead of guiding the model step by step, give it the entire problem and define the finish line up front.

👉 2. Ditch “think step by step”
Opus 5.5 handles deep reasoning natively. Old prompt hacks like “think hard” or “think step by step” only slow things down with no real quality bump. In fact, if you just want a quick reply to a straightforward question, tell it: “Answer directly.”

👉 3. Set rules of engagement in CLAUDE.md
Give the model explicit autonomy boundaries in your config file. Tell it: “When a step doesn’t need input, keep going. Stop and ask only when blocked or before anything destructive (like deleting data or force-pushing).” This prevents unnecessary pauses on routine tasks.

👉 4. Use TASKS.md to prevent context loss
Context windows get summarized during long sessions, but a persistent local file never forgets. Keep a `TASKS.md` file in your repo so Opus can log progress, check off items, and stay anchored across hours of ex*****on.

👉 5. Scale audits with subagents
Need to review a massive codebase? Fan out audits across parallel subagents, then have Opus aggregate the evidence into a single, structured summary table.

👉 6. Check pending decisions first
When a long run wraps up, don't get lost reading the full log right away. Jump straight to the open decisions or blockers that need your human input, then review the code changes.

👉 7. Handy Claude Code power tips
• Type `/fast` when doing interactive back-and-forth work to maximize token output speed.
• Drop screenshots and diagrams directly into the chat: Opus 5.5 understands layout hierarchies, connected nodes, and visual flows effortlessly.
• If a safety filter unexpectedly shifts you to an older model, quickly switch back using `/model` or hit Esc twice to rephrase your prompt.

Have you tried Opus 5.5? I'm having a great time with it.
How are you adjusting your prompting habits?

24/09/2026

A huge thank you to Packt for their continuous support of our events and community! 📚✨

Thanks to their generosity, we're always able to give away fantastic books, and our attendees couldn't be more excited.

We truly appreciate your dedication to helping our community learn and grow! 🙌

This week too we will give away 4 books at my next Google Built With AI event in Milan.
Cya there : )

16/09/2026

Sarò in Campania nei prossimi giorni e ci saranno molte occasioni per vederci:

1. NAPOLI 17 sett. sera
Domani organizziamo una pizza con qualche amico Google e del Google Group di Napoli.
Se volete partecipare, scrivetemi in pvt

2. NAPOLI 18 sett. Tutta la giornata.
Evento Google AI (masterclass + workshop). Ancora biglietti disponibili ma siamo in tantissimi.
https://gemini-codelabs-2026.vercel.app/

3. CASERTA: 18 sett. verso cena / dopo cena.
Parteciperò alla "Notte Bianca degli Innovatori"
https://www.eventbrite.it/e/notte-bianca-degli-innovatori-2-edizione-tickets-1997047306252

4. CAMPOBASSO: 19 sett. Tutto il giorno.
Ho uno speech di un'ora su UI Generativa alla Google Dev Fest Campobasso
https://gdg.community.dev/events/details/google-gdg-campobasso-presents-devfest-campobasso-2026-rebuild-your-future/

16/09/2026

Generative UI is not an LLM generating raw HTML/CSS code to blast onto the page with an innerHTML.
That is not the future of the frontend: it's just a security incident with a nice demo : )

In a Generative UI app, the model doesn't create the interface from scratch.
It produces the description of an interface using components that you have already written and typed, leaving your app with the task of rendering them.

How do we get there?

👉 BEFORE: AI TEXT CHAT
1. Users send prompts: "Who was Ada Lovelace?"
2. LLM answers with text: "Ada Lovelace was a 19th-century mathematician..."

Useful, but unstructured and non-renderable.

👉 NOW: GEN UI
1. Structured output: you pass a schema and get a guaranteed JSON every time:
{ "name": "Ada Lovelace", "role": "Mathematician", "skills": [] }.

2. Function Calling / Tools: you turn your components (e.g. ) into tools with defined parameters.
You tell the model what your app knows how to display; the model analyzes the prompt and decides which component to call and with which props.

The model handles logical routing and data extraction, your Design System handles native rendering.

Maximum generative flexibility, zero XSS, 100% control over the frontend.

15/09/2026

Lately, I’ve been getting more and more fascinated by the world of Generative UI: the idea that an interface can adapt, compose itself, or even be generated at runtime based on the user's context opens up incredible scenarios for the frontend.

Powerful frameworks and patterns are emerging: I'm thinking of hashbrown.dev (which brings this paradigm to React and Angular), the MCP apps ecosystem, or the A2UI approach.

👉 HOW TO START
The temptation to immediately dive into these high-level tools is strong, but there is a key step that cannot be skipped: understand "function calling" (or tool use).

GenUI isn't "magic": its foundations rest on the LLM's ability to map intents to structured schemas (JSON) and invoke the right logic building blocks. If you don't deeply understand how the model decides what to call, which parameters to pass, and how to manage the flow's state, using these frameworks risks turning into just a copy-paste exercise that is hard to debug.

Architectural fundamentals first, abstractions later.

👉 FUNCTION CALLING
So let's take a look at the basic operation of function calling and I'll go into more detail in the next few days

1. Your user asks something: "Is it raining in Venice?"
2. Your app calls the model with the question plus the tools you declared — each one a name, a description and a JSON schema.
3. The model doesn't answer. It replies with a tool call: {"tool": "getWeather", "args": {"city": "Trieste"}}
4. Your code runs getWeather(). HTTP call, DB query, file — whatever it really is. The model executed nothing.
5. You append the result to the same conversation and call the model again.
6. Now the model writes the answer in natural language.
7. Your app sends it back to the user.

The part people get wrong: the LLM has no access to your APIs, your DB or your machine. It only names a function and fills in the arguments.

Step 4 is the only place where code actually runs.... and it runs in your app.

👉 OTHER (REAL) EXAMPLES
In this way you can ask to your Arduino or system automation system to turn on a light, you can ask to query your databases to get the sales in 2026 of your products, access to external data and so on

In my next posts, I'll share some examples and approaches for applying GenUI to your projects.

Are you already using any tools or approaches in your projects?

02/10/2023

prossime ( in italiano e inglese)
Dopo un po' di mesi di pausa (eccetto per alcune occasioni) ricomincio a fare qualche talk con un nuovo format "interattivo".

Di seguito alcune delle conferenze a cui parteciperò come speaker e a cui non dovresti assolutamente mancare

1️⃣ Codemotion @ Milan
Language: english / italian
24-25 october
https://conferences.codemotion.com/milan2023-live/
COUPON SCONTO 20%: PARTNER_WTM

2️⃣ React Day @ Verona (the city of Romeo and Juliet) by GrUSP
Language: english
27 october: english
https://2023.reactjsday.it/

3️⃣ Google Dev Fest @ Pescara
Language: italian
18 november
https://gdg.community.dev/events/details/google-gdg-pescara-presents-devfest-pescara-2023/cohost-gdg-pescara

4️⃣ Angular Day @ Verona by GrUSP
Language: english
24 november
https://2023.angularday.it/

5️⃣ Google Dev Fest
Language: italian
https://gdg.community.dev/events/details/google-gdg-bari-presents-devfest-bari-2023/
2 december

----
Ci vediamo lì ♥️

Indirizzo

Trieste

Telefono

+393381196284

Sito Web

https://www.fabiobiondi.dev/, https://www.linkedin.com/in/fabiobiondi/, http://learnbydo.ing/

Notifiche

Lascia la tua email e sarai tra i primi a sapere quando Fabio Biondi / GDE - MVP pubblica novità e promozioni. Il tuo indirizzo email non verrà utilizzato per nessun altro scopo e potrai annullare l'iscrizione in qualsiasi momento.