KIMP’s Picks September 4th: Weekly Highlights In Design, Marketing, AI, & More

Friday’s here. Time for a quick catch-up on the biggest updates of the week, trending conversations, and AI news from the past seven days. We’ve sifted through the noise to bring you the stories, ideas, and trends worth knowing about.

Let’s get started with KIMP’s Picks’ September 4th edition.

In the Marketing Realm 

Google reminds marketers to rethink measurement as reporting gaps are growing 

Google Ads is encouraging marketers to move beyond relying on a single reporting tool when evaluating campaign performance. In its latest Ads Decoded discussion, Google highlighted that attribution, incrementality testing, and marketing mix modeling (MMM) each provide different insights and work best when used together rather than in isolation.

The company argues that attribution is useful for day-to-day optimization, incrementality studies help identify the true impact of advertising, and MMM offers a broader view of channel effectiveness over time. Combining these approaches can help marketers build a more accurate picture of return on investment.

Google also emphasized the importance of measuring the full customer journey. In addition, Google highlighted cross-channel budget planning tools in Google Analytics and continued efforts to make its open-source Meridian MMM platform more accessible through forecasting, scenario planning, and easier integrations.

These updates reflect a broader shift toward more holistic measurement that helps brands make smarter budget decisions in an ever-changing marketing landscape.

A new marketing debate in the role played by retail media in brand building 

ADWEEK recently highlighted a new debate in marketing circles which is challenging long-held assumptions about retail media. While some critics argue that retail media mainly captures shoppers who are already ready to buy, some point out that the role of retail media in brand building should be discussed more. 

The argument centers on how retail media influences both mental and physical availability. Advocates point to research showing that advertising placed close to the point of purchase can drive awareness, reach, and consideration, not just conversions. They also argue that traditional measurement models often undervalue retail media because many were built around channels like linear TV and fail to fully account for its impact.

For brands, this debate matters because it reframes retail media as a potential brand-building channel rather than just a performance marketing tool.

Sometimes ad platforms and backend data can both mislead marketers 

Another insightful marketing discussion we came across this week came from MarTech. The post discusses how some marketing platforms report much higher ROAS than a company’s backend data. But neither number necessarily reflects the true impact of advertising.

Ad platforms may count view-through conversions, modeled conversions, and older clicks within attribution windows. Backend systems often use last-click attribution, giving the sale to the final interaction, such as branded search or direct traffic. This can leave impression-led channels such as social, display, video, and CTV with little or no credit.

The problem becomes even more evident when multiple platforms claim the same purchase. A customer might see a Meta ad, later click a Google ad, and buy. Both platforms can report the same sale, while the backend records only one transaction. 

The article discusses the reality in this approach and some ways to tackle the gap. The key takeaway is to measure what advertising actually causes, instead of simply looking at which platform claims the sale. 

In the Content Marketing Realm 

LinkedIn shares new playbook for B2B brands looking to boost their AI search visibility 

LinkedIn has published a new guide focused on how B2B marketers can improve visibility in AI search as buyers rely more on AI-generated answers. The company cites research showing 94% of B2B buyers use generative AI for research, while its own marketing sites have seen non-branded search traffic fall by as much as 60%.

The guide argues that AI visibility depends on more than traditional SEO. AI systems pull information from multiple sources. Therefore, consistent coverage across brand content, employee voices, social channels, and third-party media is important.

As LinkedIn continues to be among the top cited domains for professional queries, the insights shared in this article are pretty useful for content marketers looking to revamp their strategies. 

The guide recommends building topic clusters, publishing detailed LinkedIn articles alongside shorter posts, and earning coverage beyond owned channels. It also recommends tracking AI citations, share of voice, citation position, and sentiment rather than relying only on rankings and clicks.

In the AI Realm 

Google Pics now makes image generation in Google Workspace so much simpler 

Google is rolling out Google Pics, an AI image creation and editing tool built on its Nano Banana image model. The tool is becoming available to Google AI Pro and Ultra subscribers and most Workspace business customers over the coming weeks.

Google Pics is designed for both generating new visuals and making detailed edits to existing images. The tool packs intuitive image processing features including isolating and changing individual objects, editing or translating text inside images, generating multiple versions from one prompt, and collaborating with others on image edits.

The tool will also connect directly with Google Workspace. Google has already added integration with Docs and Slides, while Drive integration is planned for the coming weeks. Users will be able to create and edit visuals without moving between separate applications.

Google introduces Gemini 3.8 with coding and reasoning-focused updates 

Google has introduced Gemini 3.8 Flash, its latest reasoning and coding model, with the same speed and introductory pricing as Gemini 3.7 Flash. According to Google, the new model improves software engineering, autonomous agent tasks, and complex multi-step reasoning.

Google says that the model performs strongly on long-horizon software engineering and specialized professional tasks, including finance and legal benchmarks.

The release also includes Gemini 3.8 Flash Cyber, a cybersecurity-focused model designed for vulnerability discovery and automated patching. It is being offered to trusted defenders through Google’s Fairwind Program. 

Google reports that the cyber model found vulnerabilities across codebases covering 20 programming languages and achieved more than 70% success on its internal benchmark. 

Meta’s making steady progress in the coding segment with the launch of Muse Spark 1.3 

Meta has released Muse Spark 1.3 with improvements aimed at agentic workflows, coding, instruction following, and long-context tasks. The model is now available through Muse Code and the Meta Model API.

The new model handles longer tasks more reliably by using tools, maintaining context, correcting gaps in its plans, and working across conflicting information. It can also ask clarifying questions, seek user input when it gets stuck, and request confirmation before consequential actions.

The update also improves multitasking and the model’s ability to recognize its own limitations, with training focused on reducing hallucinated outcomes. For coding, Muse Spark 1.3 was trained on more long-horizon engineering tasks.

Anthropic announces Claude Fable 5.1 and Mythos 5.1 models

Anthropic has released Claude Fable 5.1 and Claude Mythos 5.1, with upgrades focused on coding, knowledge work, agentic tasks, and scientific research. Fable 5.1 is generally available, while Mythos 5.1 is restricted to trusted access programs for cybersecurity and life sciences.

Fable 5.1 reportedly costs about 25% less than Fable 5 for typical workloads, thus helping developers save about 45% of the cost for highly agentic tasks. Cache reads are now priced at $0.25 per million tokens, a 75% reduction.

The company also reports major gains in scientific applications. Mythos 5.1 designed high-affinity protein binders, while Fable 5.1 created a higher-resolution elevation map covering a third of Venus. Mythos 5.1 also sped up seven open-source deep learning models by up to 2.5 times.

Runway introduces Solaris – the first interface world model 

Runway has introduced Solaris, the first model in its new Interface World Models family. The system generates interactive interfaces frame by frame instead of relying on fixed screens, templates, or conventional code.

Solaris is designed to respond directly to clicks, drags, and other user actions. This allows interfaces to change continuously and support interactions that developers may not have explicitly programmed.

The model combines a language model for reasoning with a world model for visual rendering. Runway built Solaris on its Gen-4.5 video generation technology and adapted it for real-time interaction, with a focus on visual quality at 720p.

Grok Bot arrives on Android 

Just weeks after announcing Grok Bot, xAI has announced the Android version of the tool for on-the-go AI management. 

xAI has expanded Grok Bot to Android, following its earlier rollout to desktop and iOS for eligible SuperGrok and Cursor subscribers. The app is now available through the Google Play Store and requires a supported plan. 

Grok Bot is designed as a digital teammate that can handle work across apps and inboxes while continuing tasks when users step away. Users can message Bots through a text-thread interface, run multiple Bots together, and hand off tasks without staying involved throughout the process.

Google brings agentic video understanding to Gemini 

Google has launched agentic video understanding across Gemini 3.7 Flash, 3.6 Flash, and 3.5 Flash-Lite. The feature lets Gemini actively search, scan, and inspect relevant parts of a video across frames, audio, and transcripts instead of processing footage at a fixed frame rate.

The approach reportedly can cut video analysis token consumption by up to 88%, reduce costs by up to 66%, and improve accuracy by up to 7%. The gains are especially useful for long-form content, including lectures, how-to videos, and multi-hour recordings.

Gemini can decide what to watch, how quickly to scan it, and which signal to use based on the task. This enables capabilities such as sub-second moment retrieval, anomaly detection, precise counting, and complex searches across lengthy videos.

The feature is available through the Gemini API in Google AI Studio and Gemini Enterprise Agent Platform with no additional feature fee. Google also plans to bring it to the Gemini app and YouTube’s Ask YouTube feature. 

Social Media Feature Updates and Algorithm Changes

X expands the application of X Money by moving creator payouts to the platform 

X has shifted all U.S. creator payouts to X Money. This makes the platform’s payments service the sole method for receiving earnings from both the Original Content Rewards Program and creator subscriptions.

The biggest change is speed. Creators can now access earnings instantly instead of waiting for biweekly payout cycles or meeting the previous $30 minimum threshold. U.S. creators are required to use X Money, while creators in other countries will continue receiving payments through Stripe. The move comes as X continues restructuring its creator monetization programs.

Instagram tightens rules for AI-generated profiles on the platform 

Instagram is renaming its AI creator label to AI-generated profile in an effort to make it clearer when a profile features an AI-generated person rather than a real individual.

The update follows feedback from users who said they want more transparency when interacting with accounts that appear human but are actually AI-generated. Creators who already use the existing label will be prompted to review, keep, or remove it. Instagram clarified that creators who simply use AI tools to create content do not need to apply the label.

Going forward, accounts that feature AI-generated personas are expected to add the AI-generated profile label. Profiles that do so proactively will not experience any changes to their reach.

Snapchat draws a line on full AI-generated content in Spotlight 

Similar to Instagram, Snapchat is also taking a firm step to draw a line between authentic creativity and fully AI-generated content on the platform. Snapchat is updating Spotlight’s recommendation system to prioritize original, human-created content by excluding fully AI-generated videos from recommendations.

The change is aimed at preserving Spotlight as a destination for authentic creativity as low-quality, repetitive AI-generated content becomes more common across social platforms. Snapchat believes users continue to value personal storytelling, unique perspectives, and content created by real people.

The update does not ban AI from the platform. Content enhanced or edited with Snapchat’s AI creative tools will still be eligible for recommendation, provided it includes appropriate transparency indicators.

On a Final Note 

With that, we conclude this week’s KIMP’s Picks edition. The pattern is clear – AI is changing more industries than expected and the reactions have been different too. Social media platforms that were once resisting AI-generated content are now finding ways to focus on transparency and flexibility. Staying up to date with such news week after week can be tedious. How can you change that? Sign up for The KIMP Scoop newsletter to get quick, bite-sized updates delivered to your inbox. And check out this space next week for the next edition of KIMP’s Picks.