This website uses cookies

Read our Privacy policy and Terms of use for more information.

Good day!

This week brought some notable moves from the big AI platforms, alongside a handful of genuinely new apps worth paying attention to.

From OpenAI expanding its advertising tools to new products tackling sales research, fundraising and marketing, here’s what changed — and what looks genuinely useful.

Let’s dive in!

M A R K E T I N G

🔎 LEAD SPARKER

Digest: Paste in a prospect's website and Lead Sparker searches Instagram and TikTok for untagged mentions, earned reach, competitor activity and current trends, then packages the findings into a branded pitch deck. It's doing the research before the outreach, not merely writing the email.

For: Agencies, creator-marketing teams, consultants and salespeople pitching consumer brands with an active social presence.

Solves: Spending an hour researching a prospect, collecting social evidence and manually turning it into something persuasive enough to make an outreach message feel genuinely tailored.

Price: Free. No signup required.

Setup: Easy — paste a company URL, let it build the research deck, add your own note and call to action, then export the PDF. Its usefulness will naturally depend on how much relevant activity the prospect has across Instagram and TikTok.

Try it → Lead Sparker

P R O D U C T__A N D__P R O D U C T I V I T Y

🌱 APPLYSEED

Image source: Applyseed

Digest: Tell ApplySeed about your startup once and it turns that information into answers tailored to different accelerator applications, while identifying what's missing first. The interesting part isn't AI writing — it's avoiding the same founder story being reconstructed for every form.

For: Early-stage founders applying to accelerators, incubators and startup funding programmes.

Solves: Re-entering the same company, founder, traction and market information across dozens of differently worded application forms.

Price: Free plan available with two applications per month. Pro is $29/month for 20 applications.

Setup: Easy — start with a deck, notes or a startup profile, fill any gaps ApplySeed identifies, then review its programme-specific answers before sending. One caveat is worth noticing: its programme directory currently warns that some names, dates, cheque sizes and terms are illustrative rather than verified, so funding details should be checked independently. The browser extension for typing answers directly into external forms is also marked as coming soon.

Try it → ApplySeed

P R O D U C T__A N D__P R O D U C T I V I T Y

🔥 PITCHFIRE FOR STARTUPS

Image source: Pitchfire

Digest: Upload a pitch deck and Pitchfire looks at investors' portfolios, stated criteria, investment thesis and previous reactions to startups, then tries to match the company with relevant VCs and make an email introduction. It's a useful attempt to shrink the investor spreadsheet.

For: Startup founders actively raising a seed or venture round.

Solves: Building long lists of investors, researching which firms might actually care and then trying to find a route into each one.

Price: Free.

Setup: Easy — upload your deck and Pitchfire can match and introduce you to up to three investors per day; you can also upload an existing investor list and it will attempt an introduction where it has contact data. The founder describes it as an additional fundraising channel rather than a replacement for normal outreach, which is a sensible limitation.

M A R K E T I N G

🏘️ JEVTOWN

Image source: Jewtown

Digest: Give Jevtown a headline, product listing or post and 10,000 simulated people react to it. You can inspect who stopped, liked, ignored or disliked it, along with buyer questions and price sensitivity. Interesting as a cheap pre-flight check — not as a substitute for real customers.

For: Founders, marketers, ecommerce sellers and small teams testing messaging, listings or offers before publishing them.

Solves: Having to choose between publishing an idea completely untested or spending considerably more time and money collecting real audience feedback before every small experiment.

Price: Free. No sign-in required.

Setup: Easy — enter the text you want to test and review how the simulated audience responds. The important limitation is conceptual rather than technical: these are AI-generated reactions, so a promising result should be treated as a hypothesis to test with real people, not evidence that a campaign will perform.

Try it → Jevtown

⚡ EVERYTHING ELSE IN AI TOOLS & MODELS

OpenAI expanded its advertising tools with prompt-based campaign creation in ChatGPT Work, new creative tools in Ads Manager and integrations with HubSpot and Shopify. It is also testing Sponsored Agents, which let users start a conversation with an advertiser after clicking an ad.

ChatGPT arrived in Microsoft Word, adding a sidebar for drafting from notes, summarising documents, rewriting selected text and adjusting formatting. The Word integration is available across ChatGPT plans, including Free.

xAI released Grok 4.7, a new model aimed at coding and knowledge work. It is available through the Grok API, Grok Build and Cursor, with xAI reporting improvements across several coding and reasoning benchmarks.

GitHub Copilot upgraded its code-review workflow, making it easier to distinguish unresolved issues from suggestions already addressed in later commits. Accepted suggestions can also be grouped into commits with automatically generated commit messages.

Qwen released Qwen-Image-2.1, a 7B image-generation and editing model with native transparent backgrounds, support for up to ten reference images and output up to 2K resolution. The published Qwen Research License restricts commercial use unless separate permission is obtained.

Qwen released Qwen3.8-Omni-Flash, a multimodal model that can work across text, images, audio and video within a single large context window. It can also use tools and web search, making it particularly relevant for analysing long recordings, meetings and video.

Grok released Voice Transcribe 2.0, its new speech-to-text model for recorded and live audio. It supports speaker identification, word-level timestamps, multiple audio channels and specialised vocabulary, while existing Speech-to-Text integrations can receive the upgrade without changing their code.

Perplexity Computer added effort controls, letting users choose Light, Standard, High or Ultra depending on how much reasoning a task needs. Perplexity then selects the underlying model and reasoning level automatically.

Adobe Firefly added automatic video background removal, allowing subjects to be separated from their background across an entire clip without frame-by-frame masking. The same release also expanded GPT Image 2 generation from three to ten aspect ratios.

That's it for today!

I'll be back next week to see which new launches survive the same test — genuinely new, genuinely usable and actually worth interrupting your work to investigate.

See you soon,

News Digest AI team