Changelog & monthly updates

02.26

Flo here — this month, no big headline feature. I spent February making Notis more reliable, more predictable, and harder to break. If you delegate real work to an AI, "almost works" isn't good enough.

That said, I’m also working behind the scene on a paradigm shift for Notis. 🤐 Let’s jump in.

Interactive Buttons on Slack & Telegram

Notis no longer just talks at you — now it gives you things to tap. When Notis presents options, links, or next steps, they show up as clean interactive buttons instead of raw text you have to parse and reply to.

It's a small shift that makes Notis feel less like a chatbot and more like an app you actually want to use.

Example:

  • Notis asks you to approve a draft — you tap "Approve" or "Edit" instead of typing it out.

  • Notis shares three calendar slots — you pick one with a tap.

  • A research summary lands with a "Save to Notion" button right below it.

Reliability Month

The Boring Release (a.k.a. Reliability Month)

This is the real story of February. We shipped 30+ bug fixes across the entire stack — concurrency, integrations, automations, messaging, and Notion sync.

Here's why this matters: when you tell Notis to "post this at 9 AM" or "update my CRM after every call," you need it to actually work — every time. Not 95% of the time. Every time.

What we fixed:

  • Concurrency & locks: Eliminated race conditions where parallel tasks could kill each other. Background threads no longer leak. Stalled processes get detected and cleaned up automatically.

  • Integrations & OAuth: Auth flows are faster and more resilient. Toolkit connections handle edge cases (multiple configs, first-attempt failures) without breaking.

  • Messaging: Telegram buttons no longer collide on long messages. Slack formatting is cleaner. iMessage media pipeline is more stable.

  • Notion sync: Database parsing, routing, and backfill logic are all tighter. Fewer edge-case failures during heavy usage.

We also added cost caps per agent and stalled interaction tracking — so if something does go wrong, you know about it immediately instead of discovering it three days later.

This is the kind of work that doesn't make a good demo, but makes everything else possible.

Patch Mode for Notion

Patch Mode for Notion Documents

Sometimes you don't want a full rewrite — you want a surgical edit. With patch mode enabled, Notis can update only the relevant parts of a Notion page without bulldozing everything else.

This matters most for long-living docs (plans, specs, playbooks) where stability is the whole point. You want iteration without "delete and recreate" chaos.

Example:

  • "Add a Risks section to this doc" becomes an insert where it belongs, instead of rewriting the page.

  • "Update the intro with today's context, keep the rest" stays localized to the intro.

🤖 Automation of the Month: Competitors Comparison

This month's featured automation tackles one of the most tedious tasks in content marketing — writing detailed competitor comparison blog posts. Instead of spending hours researching, formatting, and drafting, you just send Notis a competitor name and it handles the rest.

What it does:

This webhook automation takes a competitor name as input and runs a full pipeline: deep research on both the competitor and Notis (features, pricing, integrations, reviews, limitations), scrapes product illustrations from both websites, generates a custom header image, and upserts a complete, publish-ready blog post into your Notion content database — structured with a quick take, product overviews, a head-to-head comparison table, key differences, persona-based recommendations, and a CTA.

Perfect for:

  • Founders and solopreneurs who need SEO-driven comparison content but don't have a content team

  • Marketing teams that want to scale competitor pages without manual research

  • Anyone who wants a structured, fair, and visually rich comparison post — ready to publish

Watch the tutorial

🐞 Other improvements

  • Better file manipulation: A thousand tiny improvements to facilitate how Notis manipulates files between the extensions, code interpreter and it’s context window.

  • iMessage next-gen migration: the iMessage channel now supports RCS, Whatsapp and SMS - all from the same phone number.

  • Shorter onboarding: Getting to value faster with less setup friction.

  • Unified tool search: Agents find and use the right tool more consistently.

🗞️ In Other News

📰 AI News

Here's what you might have missed in AI this month:

🔥 Matt Shumer: "Something Big Is Happening" — The essay that broke the internet. 80M+ views in days. Shumer compared this AI moment to February 2020 before COVID hit, arguing we're in the "this seems overblown" phase of something much bigger. He claims he's "no longer needed for the actual technical work of my job." (fortune.com)

💰 Anthropic raises $30B and Open AI $110B — Claude Code hit $1B ARR within seven months. Meanwhile, OpenAI is targeting ~$600B in compute spend by 2030, projecting $280B in 2030 revenue. ChatGPT now has 900M+ weekly active users. The arms race is fully financialized. (anthropic.com / openai.com)

🇨🇳 The AI Cold War heats up — Anthropic accused DeepSeek, Moonshot AI, and MiniMax of "industrial-scale" distillation attacks: 24,000 fake accounts, 16M+ exchanges to extract Claude's reasoning and coding capabilities. OpenAI made parallel accusations to Congress. DeepSeek reportedly trained models on Nvidia Blackwell chips, flouting export controls. The framing from both labs: this is a national security issue, not just IP theft. (techcrunch.com)

🤖 OpenClaw / Clawdbot goes nuclear — An Austrian dev's open-source AI agent — originally named "Clawdbot" before Anthropic sent a trademark cease-and-desist — became the fastest-growing GitHub project ever (150K+ stars). It runs locally, connects to WhatsApp/Slack/email, and actually does things autonomously (reminds you of something?!). Creator Peter Steinberger joined OpenAI on Valentine's Day. Security researchers found 30,000+ exposed instances. Running it costs $300-750/mo in API tokens (still find Notis expensive ?!). It's your new weekend project, your security nightmare, AND the reason to buy that Mac Mini for local inference - or there is Notis. (ibm.com)

🧠 GPT-5.3 Codex: "The AI helped build itself" — OpenAI's Feb 5 release included a line in the technical docs that sent shockwaves: "GPT-5.3-Codex is our first model that was instrumental in creating itself." The Codex team used early versions to debug training, manage deployment, and diagnose evals. Recursive self-improvement is no longer theoretical — it's a line item in release notes. (openai.com)

🔒 Claude finds 500+ hidden vulnerabilities — Anthropic's Frontier Red Team used Opus 4.6 to scan open-source codebases and discovered over 500 security vulnerabilities that had gone undetected for decades despite expert review. They then launched Claude Code Security, which scans codebases for vulns and suggests patches — sending cybersecurity stocks tumbling in the process. (claudelog.com)

🕵️ Trump blacklists Anthropic — OpenAI swoops in hours later — Anthropic refused to lift its two red lines for the Pentagon — no autonomous weapons, no mass surveillance. Trump ordered every federal agency to cease using Anthropic and Hegseth designated it a "Supply-Chain Risk to National Security." The $200M DoD contract is dead. Hours later, Altman announced OpenAI had struck a deal for those same classified networks — claiming the Pentagon agreed to the exact same safeguards. (axios.com)

🎯 Sonnet 4.6 launches Feb 17 — Approaching human-level performance on computer use tasks (72.5% on OSWorld). Same Sonnet pricing. 1M token context in beta. Plus: Anthropic acquired Vercept (perception/interaction AI) and the Bun runtime team. The product velocity is relentless. (releasebot.io)


That's it for February. No fireworks — just a sturdier Notis.

See you next month with big news! Flo

P.S.: Record a video testimonial (in any language) and get a free month of Notis on us.


02.26

Flo here — this month, no big headline feature. I spent February making Notis more reliable, more predictable, and harder to break. If you delegate real work to an AI, "almost works" isn't good enough.

That said, I’m also working behind the scene on a paradigm shift for Notis. 🤐 Let’s jump in.

Interactive Buttons on Slack & Telegram

Notis no longer just talks at you — now it gives you things to tap. When Notis presents options, links, or next steps, they show up as clean interactive buttons instead of raw text you have to parse and reply to.

It's a small shift that makes Notis feel less like a chatbot and more like an app you actually want to use.

Example:

  • Notis asks you to approve a draft — you tap "Approve" or "Edit" instead of typing it out.

  • Notis shares three calendar slots — you pick one with a tap.

  • A research summary lands with a "Save to Notion" button right below it.

Reliability Month

The Boring Release (a.k.a. Reliability Month)

This is the real story of February. We shipped 30+ bug fixes across the entire stack — concurrency, integrations, automations, messaging, and Notion sync.

Here's why this matters: when you tell Notis to "post this at 9 AM" or "update my CRM after every call," you need it to actually work — every time. Not 95% of the time. Every time.

What we fixed:

  • Concurrency & locks: Eliminated race conditions where parallel tasks could kill each other. Background threads no longer leak. Stalled processes get detected and cleaned up automatically.

  • Integrations & OAuth: Auth flows are faster and more resilient. Toolkit connections handle edge cases (multiple configs, first-attempt failures) without breaking.

  • Messaging: Telegram buttons no longer collide on long messages. Slack formatting is cleaner. iMessage media pipeline is more stable.

  • Notion sync: Database parsing, routing, and backfill logic are all tighter. Fewer edge-case failures during heavy usage.

We also added cost caps per agent and stalled interaction tracking — so if something does go wrong, you know about it immediately instead of discovering it three days later.

This is the kind of work that doesn't make a good demo, but makes everything else possible.

Patch Mode for Notion

Patch Mode for Notion Documents

Sometimes you don't want a full rewrite — you want a surgical edit. With patch mode enabled, Notis can update only the relevant parts of a Notion page without bulldozing everything else.

This matters most for long-living docs (plans, specs, playbooks) where stability is the whole point. You want iteration without "delete and recreate" chaos.

Example:

  • "Add a Risks section to this doc" becomes an insert where it belongs, instead of rewriting the page.

  • "Update the intro with today's context, keep the rest" stays localized to the intro.

🤖 Automation of the Month: Competitors Comparison

This month's featured automation tackles one of the most tedious tasks in content marketing — writing detailed competitor comparison blog posts. Instead of spending hours researching, formatting, and drafting, you just send Notis a competitor name and it handles the rest.

What it does:

This webhook automation takes a competitor name as input and runs a full pipeline: deep research on both the competitor and Notis (features, pricing, integrations, reviews, limitations), scrapes product illustrations from both websites, generates a custom header image, and upserts a complete, publish-ready blog post into your Notion content database — structured with a quick take, product overviews, a head-to-head comparison table, key differences, persona-based recommendations, and a CTA.

Perfect for:

  • Founders and solopreneurs who need SEO-driven comparison content but don't have a content team

  • Marketing teams that want to scale competitor pages without manual research

  • Anyone who wants a structured, fair, and visually rich comparison post — ready to publish

Watch the tutorial

🐞 Other improvements

  • Better file manipulation: A thousand tiny improvements to facilitate how Notis manipulates files between the extensions, code interpreter and it’s context window.

  • iMessage next-gen migration: the iMessage channel now supports RCS, Whatsapp and SMS - all from the same phone number.

  • Shorter onboarding: Getting to value faster with less setup friction.

  • Unified tool search: Agents find and use the right tool more consistently.

🗞️ In Other News

📰 AI News

Here's what you might have missed in AI this month:

🔥 Matt Shumer: "Something Big Is Happening" — The essay that broke the internet. 80M+ views in days. Shumer compared this AI moment to February 2020 before COVID hit, arguing we're in the "this seems overblown" phase of something much bigger. He claims he's "no longer needed for the actual technical work of my job." (fortune.com)

💰 Anthropic raises $30B and Open AI $110B — Claude Code hit $1B ARR within seven months. Meanwhile, OpenAI is targeting ~$600B in compute spend by 2030, projecting $280B in 2030 revenue. ChatGPT now has 900M+ weekly active users. The arms race is fully financialized. (anthropic.com / openai.com)

🇨🇳 The AI Cold War heats up — Anthropic accused DeepSeek, Moonshot AI, and MiniMax of "industrial-scale" distillation attacks: 24,000 fake accounts, 16M+ exchanges to extract Claude's reasoning and coding capabilities. OpenAI made parallel accusations to Congress. DeepSeek reportedly trained models on Nvidia Blackwell chips, flouting export controls. The framing from both labs: this is a national security issue, not just IP theft. (techcrunch.com)

🤖 OpenClaw / Clawdbot goes nuclear — An Austrian dev's open-source AI agent — originally named "Clawdbot" before Anthropic sent a trademark cease-and-desist — became the fastest-growing GitHub project ever (150K+ stars). It runs locally, connects to WhatsApp/Slack/email, and actually does things autonomously (reminds you of something?!). Creator Peter Steinberger joined OpenAI on Valentine's Day. Security researchers found 30,000+ exposed instances. Running it costs $300-750/mo in API tokens (still find Notis expensive ?!). It's your new weekend project, your security nightmare, AND the reason to buy that Mac Mini for local inference - or there is Notis. (ibm.com)

🧠 GPT-5.3 Codex: "The AI helped build itself" — OpenAI's Feb 5 release included a line in the technical docs that sent shockwaves: "GPT-5.3-Codex is our first model that was instrumental in creating itself." The Codex team used early versions to debug training, manage deployment, and diagnose evals. Recursive self-improvement is no longer theoretical — it's a line item in release notes. (openai.com)

🔒 Claude finds 500+ hidden vulnerabilities — Anthropic's Frontier Red Team used Opus 4.6 to scan open-source codebases and discovered over 500 security vulnerabilities that had gone undetected for decades despite expert review. They then launched Claude Code Security, which scans codebases for vulns and suggests patches — sending cybersecurity stocks tumbling in the process. (claudelog.com)

🕵️ Trump blacklists Anthropic — OpenAI swoops in hours later — Anthropic refused to lift its two red lines for the Pentagon — no autonomous weapons, no mass surveillance. Trump ordered every federal agency to cease using Anthropic and Hegseth designated it a "Supply-Chain Risk to National Security." The $200M DoD contract is dead. Hours later, Altman announced OpenAI had struck a deal for those same classified networks — claiming the Pentagon agreed to the exact same safeguards. (axios.com)

🎯 Sonnet 4.6 launches Feb 17 — Approaching human-level performance on computer use tasks (72.5% on OSWorld). Same Sonnet pricing. 1M token context in beta. Plus: Anthropic acquired Vercept (perception/interaction AI) and the Bun runtime team. The product velocity is relentless. (releasebot.io)


That's it for February. No fireworks — just a sturdier Notis.

See you next month with big news! Flo

P.S.: Record a video testimonial (in any language) and get a free month of Notis on us.


Break Free From Busywork

Delegate your busywork to your AI intern and get back to what matters: building your company.

Break Free From Busywork

Delegate your busywork to your AI intern and get back to what matters: building your company.

Break Free From Busywork

Delegate your busywork to your AI intern and get back to what matters: building your company.