Olivier Bazoud

📨Email should be simple But if you've ever integrated one of the big ESPs, you know the pain. ❌ Poor developer experience ❌ Slow performance ❌ Outdated APIs ❌ Outdated UX/UI I got tired of it, so I built @nuntly_com #email #transactional #devx #api #sdk #deliverability

French gov moving faster than most companies we now have a skill to do taxes, what a time to be alive

OpenAI Backs Bill That Would Limit Liability for AI-Enabled Mass Deaths or Financial Disasters Très rassurant... wired.com/story/openai-b…

@DrizzleORM You should've announced you're switching to a Prisma-like API

After new Prisma’s drizzle-like migrations api, I don’t even know what to post on Apr 1st

"We're moving away from MCPs and instead using APIs and CLIs", Aravind Srinivas, CEO Perplexity (Ask 2026) Cloudflare confirms: MCP burns too many tokens for too little work. Where do you stand? #AI #DevTools #MCP #CLI #API #DevX

@iziatask Sympa. Y a un faux air de Superset / Conductor. C'est open source ?

J'ai supprimé Cursor, GitHub Desktop, Warp et Linear J'ai build Devflow Un env isolé par branche avec sa propre DB, ses propres ports Tout ce dont j'ai besoin est dedans

I've seen production domains silently failing SPF because nobody thought to count the lookups. It's one of those bugs that doesn't show up in any monitoring until your deliverability tanks. nuntly.com/blog/email-del… #email

Over the limit? Fixes: 1. Replace `include:` with `ip4:` for static services 2. Remove includes for services you no longer use 3. Flatten nested includes (refresh when IPs change) Nuntly uses a single include. One lookup. nuntly.com/blog/email-del… #EmailDeliverability #SPF

How to audit your SPF lookups: 1. Run `dig TXT yourdomain.com +short` 2. Count each `include:`, `a:`, `mx:`, and `redirect=` 3. Follow each include and count its nested lookups 4. Total must stay under 10 Over 10? Your SPF silently fails on every email.

Every `include:` in your SPF record triggers a DNS lookup. Nested includes count too. Google flattened _spf.google.com to raw IPs (1 lookup). Most providers still nest 2-3 levels deep. Add email platform, marketing tool, helpdesk. You hit 10 fast.

SPF records have a 10 DNS lookup limit. Most developers don't know this, and their authentication silently fails when they exceed it. Here's how to check yours and fix it: 🧵 #email #security #spf

@BrivaelFr Pourquoi tu n'as pas forké Conductor ou Superset au lieu de repartir from scratch ?

J'ai construit mon propre conductor de coding from scratch. Je pense que toutes les boîtes devraient le faire. Pourquoi les outils génériques comme Codex ou les background agents de Cursor sont frustrants : chaque boîte a une infra différente, des services différents, du logging différent. Le problème est profondément personnel. Ce qui marche beaucoup mieux, c'est un thin wrapper qui orchestre à la fois ton agent de coding et tes services internes au même endroit. C'est plus élégant, et ça débloque des choses qu'aucun outil générique ne peut offrir. Voilà à quoi ressemble la stack : Backend en Go qui gère tous les process par session. Chez Argil on a plus de 10 services, donc il gère l'isolation complète de chacun. Logs, lifecycle, tout reste propre. Le CLI est wrappé avec Tori (Rust) et j'ai ajouté des petits plugins en Swift pour avoir exactement la même expérience que Claude Code dans le terminal, y compris le drag and drop natif pour les images. J'ai pluggé un connecteur Linear. Je drop un ticket, il feed automatiquement tout le contexte dans la conversation Claude Code. Pas de copy paste, pas de context switching. Ensuite j'ai wrappé le tout avec Playwright pour que l'agent ait un vrai accès à toute l'app. Pas un mock, pas un screenshot. Le vrai produit, end to end. Le résultat : mon setup est maintenant quasi identique à ce qu'aurait un ingénieur correctement onboardé chez Argil. Même contexte, mêmes tools, même accès. Et c'est juste le début. Prochaines étapes : → Plugger d'autres types d'agents au-delà du coding. Growth, QA, deployment. Même interface, même orchestration. → Construire une decision layer custom qui va bien au-delà de "écrire du code." Un harness qui comprend le graph complet de la boîte. → Connecter tous les outils externes dans un système unifié. L'endgame d'ici quelques mois : tout le monde dans la boîte devient contributeur de ce système. Il ne reste que deux métiers : Construire et maintenir les systèmes agentiques qui font grossir la boîte Amener la taste et prendre les décisions. Tracer le context graph à l'échelle de la boîte. L'ère du "j'utilise un outil d'IA pour coder" est terminée. Le vrai levier, c'est de posséder la stack complète de comment les agents interagissent avec ta boîte.

Make your OpenClaw happy. Give it an inbox with memory! ...not only an email address. With Nuntly, your OpenClaw agent can create inboxes, receive messages, load conversation history, reply, and persist context. #OpenClaw #AIAgents #EmailAPI #DevX nuntly.com/blog/ai-agent-…

Resend gives your agent an email. Nuntly gives it an inbox with memory. That's the difference between sending notifications and having real conversations. Make your OpenClaw happy with Nuntly. #OpenClaw #AIAgents #EmailAPI #DevX

😍😍😍

Sending email is half the picture. Support tickets, billing queries, AI agents acting on inbound messages. All need a receiving API. Nuntly now handles inbound email natively. 📫 Same platform, same API, same domain. Three use cases shipped today 🧵 #inbound #outbound #DevX

@pbteja1998 Hi, how long did it take to get these? How much ?

SiteGPT.ai is now SOC 2 Type II and HIPAA compliant... SiteGPT will be 3 years old by the end of this month, and I felt like it's finally time to get SOC 2 compliant... This will open up so many new doors for SiteGPT... I am excited for it.

That blind waiting period is such a weirdly unsolved problem for how universal it is. I've definitely sat there refreshing MXToolbox wondering if I fat-fingered a TXT record or if it just needs more time. Showing propagation progress inline feels like one of those "why didn't this exist already" features

Setting up a sending domain shouldn't feel like guesswork. Every developer who's configured email DNS records knows the drill: copy a DKIM key, paste it into your DNS provider, add SPF, DMARC, MX records. Then wait 🤨 Thread 🧵

@floinkus @zenorocha We built exactly this at Nuntly. Inbound and outbound email Inbound with 1GB inbox storage. No more lost messages. nuntly.com/features/recei…

@zenorocha hey zeno -- can we store messages on Resend Receiving? we're having to migrate off it cause we lose messages :( happy to pay more

Any terminal users here? I'm working with the team on an official Resend CLI. Looking for beta users today.

One of the most underrated pain points in email infrastructure: you set up your DNS records and then... you wait. No feedback, no progress indicator, just hope. We just shipped a DNS propagation checker directly in the Nuntly dashboard. It tells you exactly what's missing.