Dutch increased conversion rates by 50% by focusing on high intent bottom-of-funnel SEO content and improving visibility in AI-driven search. It raised its LLM mention rate from 57.9% to 82.5% and improved AI search presence from 23.8% to 45.5% in two months. About 20% to 50% of traditional search traffic is at risk as roughly half of users now engage with AI-driven search. The strategy combined content expansion, technical SEO fixes, and off site efforts like backlinks and third party placements. These changes drove higher quality organic traffic and helped it compete more effectively against larger competitors.
A prioritization framework called THRICE extends the familiar ICE scoring model by adding Time, Headcount, and Reach alongside Impact, Confidence, and Effort. The additions address common SEO failures where teams burn political capital on initiatives that take too long, cost too much, or reach too few people. Each dimension is scored on a 1-10 scale, and the totals make trade-offs visible. A homepage title tag change might score 60/60 while a full site translation scores 26/60, making the case for sequencing without subjective debate. The framework also builds trust with engineering teams by openly accounting for their workload before making requests.
Google will automatically upgrade Dynamic Search Ads and campaign-level broad match to AI Max starting in September, with early migration tools available now. AI Max is not a new campaign type but a suite of features within Search campaigns that combines dynamic site matching with broader AI-driven query matching, text optimization, and landing page selection. Advertisers who used DSA as a catch-all discovery layer will need to shift toward more deliberate campaign structures since AI Max rewards clean site architecture and strong landing pages over manual keyword builds.
This webinar takes place today at 9 AM PT. Leaders from Optimizely, CompTIA, and 3Play Media break down their tools and team structures. They will share how they train teams and build AI fluency. Attendees will learn how to integrate AI, whether they are just starting or scaling.
AI agents now act as buyers, and product pages must be optimized for them to increase selection rates. New research from Yale and Columbia finds that small changes in wording can drive large gains. The study found that renaming a product to match intent increased selection by 80.4 percentage points for GPT-5.1, 52 points for Gemini 2.5 Flash, and 41 points for Claude Opus 4.5. Agents favor strong keywords, higher ratings, more reviews, and positive badges, while “Sponsored” labels reduce selection. Better models make fewer mistakes, as failure rates dropped from 63.7% to 4.3% for Claude and from 25.8% to 1% for GPT across versions.
Influencer marketing leads at Duolingo and Crocs shared lessons from running major creator campaigns. A Duolingo partnership that started as a cold DM two years before activation generated millions of views during Miami Music Week, illustrating how patience and relationship-building produce results that transactional approaches cannot. Both marketers stressed that open briefs consistently outperform scripted ones because creators build audiences through their own voice, not brand talking points. A Crocs dark romance parody driven by a team member's genuine passion for BookTok earned 14 million organic views and eventually led to a legitimate Twilight franchise collaboration.
An Austin-based window cleaning franchise has turned a commoditized home services business into a lifestyle brand through bold retro branding, a distinctive color identity, and uniforms that double as sellable merchandise. The company operates over 100 franchise locations and positions its marketing infrastructure as a core selling point for franchisees, providing SEO, campaigns, and creative assets as part of the package. Branded vans parked at data-selected high-traffic locations replaced traditional billboard spending while building rapid local awareness.
Rising memory costs driven by AI demand are pushing smartphone prices up by 40% to 50% and forcing major brands to cut output and rethink products.
2026.04.17·
cafetechinenglish.substack.com·2 min read·TLDR Marketing
Sponsored
Your customers expect your product or AI agent to work with the tools they already use (like Slack or Salesforce). But every integration your team builds means owning the auth, the webhooks, the schema changes, and the maintenance indefinitely.
Platforms are not just products - they are environments that shape how teams grow and operate. Success comes from building a system that makes teams more capable, not just more compliant.
Even when dashboards are embedded, usage still drops off. The issue isn't access, it's action. Analytics show what's happening but leave users figuring out what to do next. In this blog [10], Reveal breaks down why that gap kills adoption and what teams are doing to close it.
Product success now depends on building to learn before building to earn. The best teams prototype rapidly to discover what works, rather than just shipping faster.
AI credit systems need hierarchy, real-time enforcement, and clear visibility. The best designs unify decisions, track usage, and let enterprises control spend across teams.
2026.04.17·
thefinancialengineer.substack.com·10 min read·TLDR Product
NPX WORKOS@LATEST launches an AI agent, powered by Claude [5], that reads your project, detects your framework, and writes a complete auth integration into your codebase. No signup required. It creates an environment, populates your keys, and you claim your account later when you're ready.
Elon Musk's Terafab team has reached out to chip industry suppliers for price quotes and delivery times for a variety of chipmaking gear. The outreach suggests Musk is continuing with the project in the face of scepticism from the semiconductor industry. Musk's representatives have asked for speedy price estimates while providing minimal information about the products to be made. The project is offering to pay a considerable amount above quoted figures if suppliers give Terafab priority.
Apple is sending a large portion of its Siri engineers to a multi-week bootcamp on learning how to code with AI. Around 60 members of the Siri development team will continue working on Siri, with another 60 evaluating how the assistant is performing. Apple is expected to unveil a smarter, more capable version of Siri in two months at WWDC. The new Siri is under testing to ensure it meets Apple's safety standards and can interpret and execute commands from users.
The Artemis III mission next year will help NASA and its contractors set up an attempt at one or more visits to the moon in 2028. Both SpaceX and Blue Origin have run into delays in developing spacecraft for the missions. A landing operation with astronauts within the next few years may not be achievable. Both companies have key tests coming up that will inform their lunar work and vehicles.
The paradigm is shifting to robot learning, autonomous science, and human-machine interfaces. These are not entirely separate efforts and thematically form a group of emerging frontier systems for the physical world. They share common technical primitives and are mutually reinforcing in ways that create compound dynamics across domains. These domains create a structural flywheel for extending AI into the physical world. This post looks at the technical primitives underlying these domains and examines why they present frontier opportunities.
DigitalOcean Deploy is April 28 in SF. Leaders from Character, Workato, VAST Data, Arcee, and vLLM on stage sharing how they run AI in production. Plus a fireside chat with NVIDIA's Kari Briski on the future of agentic AI and a first look at DigitalOcean's next-gen
OpenAI has introduced a number of new features designed to help businesses create their own agents using its agents software development toolkit. This includes a sandboxing ability that allows agents to operate in controlled computer environments. There is a new in-distribution harness for frontier models that allows agents to work with files and approved tools within a workspace. The new features are designed to allow users to build long-horizon agents on their own infrastructure.
Xata is a cloud-native Postgres platform with fast branching, scale-to-zero functionality, 100% vanilla Postgres, and all the production-grade requirements. It can be used to create preview and testing environments quickly and with real data, and as a fully-managed PostgreSQL service to run the production database. The core Xata platform is open-source under the Apache 2 license.
Snap plans to cut around 1,000 jobs from its full-time workforce and close more than 300 open positions. The company is facing pressure from investors over its struggles for profitability. The cuts will enable Snap to increase efficiency and pursue profitable growth. Snap now projects that its total revenue will increase by 12% in the first quarter.
ChatGPT is an aggressive editor. It favors its general search index, uses semantic similarity to select and cite sources, and treats Reddit like a textbook. The pages that get cited are the ones whose titles and content match the questions ChatGPT is asking behind the scenes, and that surface through the right retrieval channel.
The US Office of Science and Technology Policy has published an initiative to develop low- to mid-power space reactors in orbit and on the lunar surface.
TLDR's Applied AI team is tasked with making every process at TLDR legible to code, runnable by anyone, and composable into larger workflows. Join a small, fast moving team using the latest AI tools with an unlimited token budget. Learn more [21].
Usage, credits, access, GBB… which monetization model makes sense for AI? Metronome's Pricing Index covers 50+ AI companies, and they've shared some takeaways in a new blog [4].
Google's Gemini 3.1 Flash TTS enhances text-to-speech with improved expressivity and controllability, featuring a notable Elo score of 1,211 on the Artificial Analysis TTS leaderboard. The model supports over 70 languages and introduces audio tags for granular control of vocal style, allowing easy manipulation via natural language commands. All generated audio is watermarked with SynthID to ensure authentic content, preventing misinformation.
OpenAI introduced updates to its Agents SDK, adding a model-native harness for cross-file and tool workflows along with sandboxed execution for safer task handling.
Humwork launches the first Agent-to-Person (A2P) marketplace to connect AI agents with verified human experts when AI tools encounter challenges. The platform integrates with AI-centric tools like Claude Code and Replit, allowing handoffs to occur in under 30 seconds with full session context shared securely. With more than 1,000 experts available globally, Humwork boasts an 87% resolution rate and is backed by Y Combinator's P26 batch.
IBM Research uses an executable benchmark with thousands of APIs and documents to test multi-step agent reasoning and tool use, revealing consistent performance gaps and common failure modes.