Ciente > Articles by: Ciente Editorial Team

DeepSeek’s V4-Flash Is Unreasonably Cheap, and That’s Bad News for US AI Margins

Author: Ciente Editorial Team

DeepSeeks
News

DeepSeek’s V4-Flash Is Unreasonably Cheap, and That’s Bad News for US AI Margins

DeepSeek’s V4-Flash Is Unreasonably Cheap, and That’s Bad News for US AI Margins

DeepSeek’s new V4-Flash costs pennies compared to OpenAI and Anthropic models. Could ultra-low inference costs be changing the economics of enterprise AI?

DeepSeek is back doing the one thing that terrifies Silicon Valley CFOs: making Western AI look wildly overpriced.

The Chinese startup rolled out V4-Flash, its latest low-cost model, on Friday.

According to research firm Artificial Analysis, running a standardized test suite on V4-Flash costs about 3 cents. Running that same test on OpenAI’s GPT-5.6 Sol costs $1.86. On Anthropic’s Claude Fable 5, it costs $3.15. That is a hundredfold price collapse.

Now, let’s talk about the actual performance. V4-Flash won’t win awards for deep reasoning. It scored 50 out of 100 on Artificial Analysis’s Intelligence Index, placing it on par with Google’s Gemini 3.6 Flash and well behind frontier models. If you need an AI agent to architect complex software or solve heavy logic problems, you still pay the premium for top-tier models.

But most enterprise AI work isn’t complex logic. It is mundane plumbing: sorting support tickets, extracting data from invoices, or summarizing internal notes. A “50 out of 100” score works just fine for those routine tasks. When a model gets the job done, corporate buyers stop caring about benchmark bragging rights and start looking at their API bill.

That is where DeepSeek creates real market pressure.

The startup, reportedly prepping for an IPO, doesn’t need to beat OpenAI on raw intelligence. It just needs to bleed off the high-margin enterprise traffic that funds American AI labs. When everyday tasks cost virtually zero, convincing a CFO to pay $3 per test becomes a tough sell.

OpenAI and Anthropic built incredible reasoning engines, but DeepSeek just reminded everyone that in business, price per task usually wins.

Customer Data
Blogs

Top Customer Data Platforms in 2026: What Nobody Tells You Before You Buy One

Top Customer Data Platforms in 2026: What Nobody Tells You Before You Buy One

Marketers have been promising a 360-degree customer view for a decade now. But even today, most CDPs still can’t deliver it. So, what precisely separates the top customer data platforms from expensive new silos?

The 360-degree customer view has been the selling point of every CDP pitch for the last decade.

It sounds irresistible. One unified profile per customer. Every touchpoint, every purchase, every interaction, all in one place. Marketing finally knows what sales knows. Product finally sees what support sees. Personalization becomes genuinely personal.

Most implementations don’t get anywhere near that.

Not because the technology is bad. Because the data going into these platforms is fractured, inconsistently structured, and often years of technical debt layered on top of more technical debt. A CDP doesn’t fix broken data infrastructure. It exposes it faster and at greater cost.

That’s the uncomfortable truth the vendor demo never covers. And it’s the reason so many companies buy a top customer data platform, spend six months on implementation, and end up with a shiny new silo sitting next to all the old ones.

Knowing which platform to buy is maybe 20% of the problem. Knowing what you actually need from one is the other 80%.

What a Top Customer Data Platform Actually Does

Strip away the marketing language and a CDP does three things.

1. First, it ingests data from every relevant source: websites, apps, CRMs, email platforms, POS systems, support tools, offline channels. Everything that generates a signal about a customer flows into one central location.

2. Second, it resolves identity. This is the hard part. A customer browsing on mobile, purchasing on desktop, calling support from a phone number, and opening emails from a work address looks like four different people without identity resolution. A good CDP stitches those interactions into one coherent profile. A weak one leaves them fragmented and calls it a “unified view” anyway.

3. Third, it activates. Unified profiles feed into segmentation, personalization, campaign automation, and downstream analytics tools. Strong customer activation strategies ensure unified data translates into measurable business outcomes instead of sitting in reports. This is where the investment either pays off or disappears into a reporting dashboard nobody checks.

The platforms that do all three well are rare. Most excel at one or two and paper over the gaps.

The Two Types of Top Customer Data Platforms (And Why the Distinction Matters)

The market splits cleanly into two categories. Most buyers don’t realize this until they’re deep into implementation.

Actionable Customer Data Platforms

These are built for marketers. They unify data and also provide native tools to activate it, cross-channel campaign automation, personalization engines, journey builders, push notifications, SMS, email, WhatsApp. They are especially effective when paired with proven customer lifecycle management practices that engage users beyond a single campaign. The data and the action live in the same platform.

Insider One, Bloomreach, Salesforce Marketing Cloud CDP, Adobe Real-Time CDP, Emarsys, and Optimove fall into this category. The appeal is obvious: fewer integrations, one interface, faster time to value for marketing teams. The tradeoff is flexibility.

Actionable CDPs are optimized for specific use cases, and complex data science work often still needs to live elsewhere.

Access and Analytics Customer Data Platforms

These are built for data teams.

Twilio Segment, Tealium, mParticle, Treasure Data, and ActionIQ prioritize data collection, governance, and routing into existing infrastructure like Snowflake or BigQuery. They don’t dictate how you activate the data. They get it clean and accessible so the tools you already use can do the activation.

The advantage is architectural flexibility. The disadvantage is that marketing teams often can’t use them without engineering support. Every personalization use case requires an integration. That’s fine for companies with mature data teams. It’s a roadblock for everyone else.

Knowing which category fits the team doing the actual work is the first decision. Not the last one.

What Actually Separates the Top Customer Data Platforms

Identity Resolution Quality

Everything downstream depends on this. If a CDP builds profiles out of duplicate records and mismatched identifiers, every segment, every personalization, every insight built on top of it reflects a fictional version of the customer. Accurate identity resolution also strengthens customer experience analytics by ensuring insights are based on reliable customer data.

Amperity built its reputation on this specific problem. Its machine-learning-powered identity resolution handles messy, legacy, enterprise-scale data better than most competitors. For retailers and hospitality brands sitting on decades of fragmented customer data across dozens of systems, that matters enormously.

The key questions to ask any CDP vendor: What methodology drives your IDR? How do you handle conflicting identifiers? What does the match rate look like on messy legacy data, not clean demo data?

Real-Time vs. Batch Processing

Personalization that fires six hours after the triggering event isn’t personalization. It’s a belated reaction. The psychology of personalization shows that timely and relevant interactions have a much stronger impact on customer engagement.

Adobe Real-Time CDP processes data in milliseconds. It handles streaming event data at scale, which matters for brands where the customer decision window is measured in seconds, not hours.

Not every company needs this level of speed, but the ones that do, e-commerce brands, streaming platforms, financial services, need to know whether the CDP can actually deliver on the “real-time” claim in their specific infrastructure context, not just in principle.

Composable Architecture vs. Packaged CDP

This distinction is reshaping the market fast.

Traditional packaged CDPs move all your data into their proprietary storage. That creates a new silo. You pay for storage you already own elsewhere, compliance becomes harder to audit, and switching costs grow every month.

Composable CDPs sit on top of your existing cloud data warehouse. Snowflake, BigQuery, Databricks. The data stays where it is, governed as it always was. The CDP provides the identity resolution, segmentation, and activation logic on top. DataOS takes this furthest, treating customer data as reusable, governed data products with built-in lineage and access controls.

For regulated industries, healthcare, finance, any environment where data governance isn’t optional, composable architecture isn’t just more elegant. It’s often the only viable option.

AI Readiness

Not the AI the vendor built into their platform. The AI your team wants to bring.

Can the CDP host and run custom machine learning models directly on unified customer profiles? The ability to analyze unified customer data efficiently also supports broader data analytics initiatives that improve customer experience. Or does every predictive use case require exporting data to a separate environment, running the model there, and reimporting the results? That workflow sounds manageable. At scale, with multiple models refreshing constantly, it becomes a serious operational bottleneck.

Treasure Data and ActionIQ both handle complex AI and ML workloads well. Insider One builds predictive scoring natively, things like purchase likelihood, churn risk, discount affinity, directly into the profile layer without requiring a separate data science pipeline. Different approaches. Both solve real problems for specific team structures.

The Top Customer Data Platforms Worth Knowing in 2026

A. Twilio Segment still sets the standard for developer experience. 700-plus pre-built connectors, clean event tracking APIs, strong data governance via Segment Protocols. The right fit for mobile-first B2C teams who need data flowing reliably across a best-of-breed MarTech stack.

B. Salesforce Data Cloud makes the most sense for companies already deep in the Salesforce ecosystem. The native integration with Sales Cloud and Service Cloud creates coordination between marketing, sales, and service that’s genuinely difficult to replicate across separate platforms. The tradeoff is lock-in, and the price reflects the premium of that coordination.

C. Adobe Real-Time CDP earns its place for global B2C brands running complex, high-volume personalization across digital channels. Adobe Sensei powers genuinely useful predictive features. The implementation is heavy and the cost is enterprise-grade, but for brands where digital experience is a competitive differentiator, the capability justifies the investment.

D. Tealium sits neutrally above complex multi-vendor MarTech stacks. It’s the best choice for teams that prioritize real-time consent management and zero-party data governance without wanting to rebuild their existing vendor relationships.

E. Amperity is the specialist pick for identity resolution at enterprise scale. If fragmented, messy legacy data is the primary obstacle, no platform handles it more reliably.

F. Insider One leads for e-commerce and B2C retail teams who want unified data and cross-channel campaign activation in one place. Native automation capabilities also help organizations improve marketing automation ROI by reducing manual effort while delivering personalized experiences. The time-to-value is fast because implementation includes hands-on data setup from the vendor’s team, not just documentation and a support queue.

G. ActionIQ works best for large enterprises with significant data warehouse investments who want to activate that data without moving or duplicating it. The HybridCompute architecture lets segmentation logic run where the data already lives.

The Customer Data Platform Mistakes That Kill Implementations

Buying architecture before fixing data quality. A CDP running on unresolved duplicates, inconsistent field naming, and broken event tracking produces confident-looking output built on unreliable inputs. Garbage in. Organized garbage out.

Treating implementation as a technical project instead of an organizational one.

The data teams, marketing teams, product teams, and leadership all have different definitions of what a unified customer profile should contain. Those disagreements surface during implementation and stall everything. The conversation has to happen before the contract is signed.

Measuring success in profiles created instead of decisions improved. A CDP that produces 50 million unified profiles means nothing if the marketing team still builds segments the same way they did before. Establishing a clear voice of customer program helps organizations measure whether unified customer data is improving real business decisions.

Ignoring compliance from the start. GDPR, CCPA, and the expanding state-level privacy legislation in the US mean consent management and data lineage aren’t optional features to bolt on after launch. Any CDP that requires a separate tool for governance is adding compliance brittleness, not removing it.

Choosing the Right Top Customer Data Platform for Your Business

The evaluation criteria that matter are simpler than most vendor comparison guides suggest.

Who does the work? If the primary users are marketing teams, an actionable CDP with native campaign tools will deliver faster value than a data engineering-first platform that requires six integrations to run a personalized email. The right platform also makes it easier to execute effective lead nurturing strategies using unified customer data. If the primary users are data and analytics teams, the opposite is true.

What does your data actually look like? Clean, well-structured, consistently tagged event data opens up more platform options. Messy, legacy, multi-system data narrows the shortlist toward platforms with stronger identity resolution capabilities.

Where does your data live? Companies invested in Snowflake or Databricks should evaluate composable architecture first. Companies without a mature data warehouse can let the CDP provide that layer without paying twice for storage.

What does success look like in twelve months? Not in five yeaWhat does your data actually look like? Clean, well-structured, consistently tagged event data opens up more platform options. Messy, legacy, multi-system data narrows the shortlist toward platforms with stronger identity resolution capabilities.rs. The realistic near-term outcome shapes which platform features matter and which are just expensive extras on a vendor slide.

The top customer data platforms in 2026 are genuinely capable. None of them replace the organizational work of agreeing on what customer data should mean, who owns it, and what the business intends to do with it. The platform is infrastructure. The strategy has to exist first.

Linkedin
News

LinkedIn Finally Gives You a Button to Flag AI Slop

LinkedIn Finally Gives You a Button to Flag AI Slop

LinkedIn quietly added a “Seems like AI slop” button to post menus. The crowdsourced spam detection reveals a bigger problem for the professional network.

LinkedIn just gave users the exact tool they wanted: a button to call out AI garbage.

You can now tap the three dots on any post and select “Seems like AI slop.” The app immediately hides the post and sends feedback straight to LinkedIn’s feed algorithm.

The update tackles a massive problem. Research firm Pangram found that AI generates over 40% of long-form posts on LinkedIn. In fact, LinkedIn hosts nearly two-thirds of all AI text across social media. Feeds that once featured real career advice are now flooded with fake inspirational stories, repetitive bullet lists, and automated comments.

LinkedIn’s Chief Product Officer Hari Srinivasan announced this on Thursday. He admitted that static filters struggle to define “slop” because low-quality content constantly changes shape.

By letting real people report unnatural posts, LinkedIn aims to retrain its detection models more quickly. The company also killed its aggressive “enhance with AI” drafting button. Writers now get a simple proofreader that fixes typos without rewriting their personal voice.

Yet, massive irony remains.

LinkedIn still pushes Premium AI writing tools to the exact users creating this fluff. A platform cannot hand everyone a text generator and then act shocked when feeds turn into a ghost town of machine-written posts.

Crowdsourcing content reporting gives annoyed professionals a small win. But user flagging will not cure LinkedIn’s deeper addiction to cheap engagement. Until social networks stop rewarding low-effort posting, users will keep pressing that slop button.

Apple
News

Why Is Wall Street Freaking Out if Apple Just Crushed Earnings?

Why Is Wall Street Freaking Out if Apple Just Crushed Earnings?

Apple blew past revenue targets with $109 billion in sales, but just hours later supply shortages sent the stock sliding. Is this what the panic is all about?

Apple just pulled off its biggest June quarter in company history. Consumers snapped up iPhones and Macs despite rising prices across the tech sector, driving total revenue to $109.4 billion- up 16% from last year.

iPhone sales jumped 22% to $54.2 billion, setting a summer record. Mac sales surged 29% to $10.4 billion- propelled by strong demand for new MacBooks. Meanwhile, profits hit $2.02 per share- topping Wall Street expectations.

The traders still immediately dumped the stock, with shares decreasing 6% in after-hours trading.

Why the sudden panic?

Wall Street fixated on supply chain bottlenecks. Outgoing CEO Tim Cook warned that global memory chip shortages are throttling production. Because Apple cannot build devices fast enough to meet demand, CFO Kevan Parekh projected 9% to 11% growth for next quarter- slightly below Wall Street’s 12% estimate.

Punishing a company for selling products faster than factories can produce them misses the mark. Apple’s real story isn’t a weak forecast. It is relentless consumer demand. Buyers are upgrading devices even as economic headwinds force price increases elsewhere.

This quarter also marked Tim Cook’s final earnings call as CEO before handing leadership to John Ternus. Cook leaves Apple with a record 2.5 billion active devices and $30.7 billion in quarterly Services revenue.

Short-term supply shortages will clear up, but Apple’s massive market dominance isn’t going anywhere.

Revenue Intelligence
Blogs

Understanding Revenue Intelligence: What Does It Mean If Not More Data?

Understanding Revenue Intelligence: What Does It Mean If Not More Data?

Revenue intelligence promises smarter GTM decisions. Most B2B teams are running it on data that’s 30% wrong. What actually separates insight from noise?

B2B GTM teams might have a misdiagnosed data problem- they

 still believe they need more data to improve their tactics.

More contacts, more intent signals, more platforms feeding more dashboards. So they buy more. Stack more tools. Build more reports. And somehow, the pipeline doesn’t get more predictable. The reps don’t get more efficient. Even the outreach doesn’t get any more relevant.

So, what’s truly happening here?

The problem was never volume. It was quality, context, and what the team does with the signal once it surfaces. That’s the entire premise of revenue intelligence done properly, much like a strong market intelligence framework that helps businesses turn data into strategic decisions. Not a bigger database. A more intelligent system that translates what you already know into decisions that move revenue.

And most B2B teams haven’t built that system yet.

What Revenue Intelligence Actually Means for B2B GTM Teams

Revenue intelligence is the discipline of converting raw commercial data into decisions that directly affect pipeline, conversion, and retention.

That definition sounds broad because the function actually is broad. It influences every part of the GTM motion:

  • Which accounts must be prioritized this week?
  • Which deals carry real momentum versus false confidence?
  • Which messaging is landing with which personas?
  • Which reps need coaching on which part of the sales cycle?

All of it runs on the same underlying logic: collect the right data, analyze it in context, and put the finding in front of the person who can act on it before the window closes.

What it isn’t is a reporting function.

Revenue intelligence that produces weekly summaries for leadership to nod at before moving on to the next agenda item is just expensive documentation. Like effective business intelligence, the function earns its value when it changes a decision, not when it describes what already happened.

Why Revenue Intelligence Fails Without the Right Data Foundation

Before any analysis happens, the data underneath the system has to be worth analyzing. This is where most revenue intelligence programs hit a wall they didn’t see coming.

Data Decay and the Revenue Intelligence Blindspot

B2B data decays fast. Contacts change roles. Companies restructure. Decision-making units shift. Research puts average data decay across B2B contact databases at roughly 30% annually. Which means a list that was reliable twelve months ago has one in three records pointing somewhere wrong today.

A rep calling a stale number isn’t just wasting time. They’re burning a touch on an account that might actually be in-market, with the wrong contact, at the wrong number, with messaging calibrated for a role the person no longer holds. That’s not a reach problem. That’s a data quality problem that looks like a performance problem until someone traces it back to the source.

Revenue intelligence built on decaying data produces confident-looking outputs rooted in a reality that no longer exists. The model scores the account highly. The rep reaches out. Nobody picks up. The cycle repeats while the team debates why the sequence isn’t converting.

The Contactability Problem Revenue Intelligence Programs Consistently Underestimate

Contactability is the gap between having a name and being able to actually reach the person attached to it.

Most data providers measure coverage. How many contacts, how many companies, how many industries. Fewer measure how many of those contacts actually pick up a phone or respond to an email from an address that exists and doesn’t bounce.

That distinction is enormous in practice. A list of 50,000 contacts with 40% contactability produces fewer real conversations than a list of 15,000 contacts where 85% of the records connect.

Revenue intelligence only runs on conversations. Without contactability, the data layer produces reach attempts, not engagement. And pipeline doesn’t come from reach attempts. It comes from conversations with the right people at the right moment.

How Revenue Intelligence Connects Data to GTM Decisions

Conversation Intelligence as a Revenue Signal

Conversation intelligence is one of the most underleveraged inputs in modern revenue intelligence programs, especially when paired with AI deal intelligence to uncover patterns that improve sales execution.

Every call a rep makes is data. Not just whether it connected, but what the prospect said, how they responded to specific messaging, which objections surfaced, which questions they asked, and at what point the conversation shifted or stalled.

Aggregate that across hundreds of calls and the patterns that emerge tell you things no intent platform can surface. Which value drivers resonate with which personas. Which objections signal genuine hesitation versus negotiation tactics. Where reps consistently lose control of the narrative.

The teams that use conversation intelligence well don’t treat it as a call recording archive. They treat it as a coaching and messaging feedback loop. The rep who learns that a specific reframe lands consistently in a certain vertical adjusts their approach in real time. The marketing team that sees a competitor being mentioned in 40% of calls in a specific segment updates the battlecard before the next quarter’s pitch.

That’s revenue intelligence working the way it should. Not documenting what happened. Feeding forward into what happens next.

Deal Intelligence and What It Reveals About Pipeline Health

Most pipeline reviews are optimistic by default. Deals stay green until they suddenly go red. Nobody catches the drift between those two states because the signals are subtle and spread across too many platforms for anyone to track manually.

Deal intelligence solves this. It monitors account engagement across touchpoints, tracks whether the right stakeholders are involved at the right stage, flags deals where momentum has slowed without anyone registering it consciously, and surfaces the patterns that differentiate deals that close from deals that slip.

The most common finding when teams run this properly is sobering.

A meaningful slice of the deals categorized as “late stage” show zero multi-threaded engagement. One contact. One relationship. No other stakeholders touched in thirty days. That deal isn’t late stage. It’s fragile. And without deal intelligence surfacing that fragility, the rep and the manager miss it until the contact goes quiet and the deal disappears from the forecast.

What Makes Revenue Intelligence Actionable Across GTM Teams

The data can be clean. The analysis can be accurate. The revenue intelligence program still fails if the insight doesn’t reach the person who can act on it, in time to act on it.

Distribution is the operational problem most programs don’t solve. A finding that goes into a weekly report gets read on Friday afternoon, processed superficially, and forgotten by Monday morning. A finding that surfaces as an alert in the rep’s CRM before a Tuesday call gets acted on in the next 24 hours. Same information. Completely different outcomes.

Sales and marketing operating from separate data sets is the other failure mode.

Marketing runs campaigns against a broad ICP list. Sales prioritizes a different set of accounts based on a different scoring model. Neither team knows what the other is seeing. The account that marketing is warming up with content gets a cold outreach from sales with no reference to any prior engagement.

The buyer experiences that as disjointed, and it damages the credibility the content work was building.

Revenue intelligence becomes genuinely valuable when every function in the GTM motion works from the same scored, current, contextualized view of the market. That requires integration. Not just between tools, but between teams and the data each team trusts.

AI and Revenue Intelligence: Where It Helps and Where It Doesn’t

AI has changed the economics of revenue intelligence meaningfully. Tasks that used to take analysts days now take minutes. Account profiling, intent signal aggregation, deal scoring, call summarization, pipeline risk flagging, all of it processes faster with AI in the stack, reflecting how artificial intelligence continues to reshape modern business functions.

But AI in revenue intelligence has a ceiling that’s often undersold in vendor conversations. The ceiling is data quality.

An AI scoring model trained on clean, validated, continuously refreshed data surfaces genuine signal. The same model trained on a database with 30% decay, inconsistent field mapping, and no tele-verification layer produces high-confidence outputs built on bad inputs. The confidence makes it worse, not better.

Teams act on the score without questioning the data underneath it, and the decisions look smart until the results come back.

The balance that works is AI for speed and pattern recognition, human validation for accuracy and usability. AI processes the volume. Human tele-verification confirms the contactability, the role, the account situation. Together, they produce intelligence a GTM team can actually rely on without running a manual audit every time a high-priority account surfaces, which is the foundation of successful business intelligence for marketing.

Building a Revenue Intelligence System That Scales Without Breaking

Start with the data layer. Not the analytics platform, not the AI tool, not the dashboard. The data.

Then define contactability standards before anything else. What does a usable contact record look like? Verified direct dial, confirmed email, current role, mapped to the right account hierarchy. Set that as the baseline. Treat anything that doesn’t meet it as incomplete, not as a lead.

Then build the feedback loop.

Every outreach that connects or doesn’t, every deal that closes or slips, every conversation that surfaces an objection or confirms a value driver, feeds back into the data model. The system improves because the team teaches it what good looks like in their specific market, with their specific ICP, against their actual closed-won patterns.

Then solve the distribution problem. Not with a dashboard. With alerts, triggers, and integrations that put the right finding in front of the right person at the right moment, similar to how modern business intelligence platforms deliver actionable insights across teams.

A revenue intelligence system that requires people to go looking for insights is a system most people won’t use consistently.

Revenue intelligence at scale doesn’t look like a bigger database or a more sophisticated analytics layer. It looks like a GTM team making faster, more confident decisions with fewer blind spots. The data makes that possible. The system makes it repeatable.

And the discipline of treating insight as an operational input rather than a reporting output is what makes revenue intelligence compound over time.

Meta
News

Meta’s New AI Algorithm Reads Your Mood to Keep You Scrolling

Meta’s New AI Algorithm Reads Your Mood to Keep You Scrolling

Meta uses LLMs to analyze the tone of every Instagram post, but users have grown skeptical. Would its AI upgrade be secretly boosting screen time?

You aren’t the only one if you caught yourself scrolling on Instagram longer than intended this week.

Users’ time on Instagram jumped double digits YoY according to Meta in its Q2 earnings call with investors. The driver behind that spike isn’t a new visual redesign. It is a quiet overhaul of the underlying recommendation algorithm.

Meta now runs every public Reel and feed post through LLMs before serving it to users. Older algorithms merely tracked your likes and watch times. The new system reads what a post actually means, i.e., its topic, tone, and context, and matches that nuance against your personal viewing history.

From an engineering standpoint, this is a brilliant technical pivot. Meta’s AI actually understands content rather than relying on crude signals like clickbait. The system delivers remarkably accurate recommendations, and a single Reels update boosted overall user sessions by 15 basis points.

This technical achievement cuts both ways.

The same algorithm that delights users has also handed state prosecutors fresh ammunition.

Dozens of US states are currently suing Meta, with claims that the tech powerhouse designs its apps to trap young users in endless scrolling loops. And releasing a smarter AI engine that measurably increases screen time validates those claims in court.

Still, Meta shows no signs of pulling back. The company spends billions on AI infrastructure because personalized feeds directly drive ad revenue. Meta built an algorithm that understands human interest better than ever, and users simply cannot look away.