TL;DR

• 2026 NFL Week 4 takeaways start in Cleveland: the Browns’ rookie WR duo Denzel Boston and KC Concepcion didn’t just flash—they carried 41% of Cleveland’s total yardage in a 27–24 win sealed by Andre Szmyt’s 56-yarder with 0:10 left. [2][3][4]

• Pittsburgh’s problem isn’t 42-year-old Aaron Rodgers’ arm; it’s oxygen. Against 73.3% man coverage he rarely saw clean windows, and the WR room beyond DK Metcalf still isn’t scaring anyone. [6][10]

• The Browns’ post–Myles Garrett identity is here: Monken-ball on offense, youth everywhere, and special teams willing to cash in from 56. The Mason Graham injury now forces Andrew Berry to find trench help to keep this formula afloat. [2][7][8]

What the source said

Bleacher Report’s Week 4 table noted five undefeated and five winless clubs entering the weekend, with Thursday night delivering Browns 27, Steelers 24 at Cleveland Browns Stadium on September 30, 2026. The piece spotlighted rookies KC Concepcion and Denzel Boston in a revamped Cleveland offense and flagged a potentially pivotal injury to second-year DT Mason Graham. On Pittsburgh’s side, it said Aaron Rodgers is getting little help from a receiver group that can’t separate, while RB Jaylen Warren remains a bright spot. It also previewed Monday’s Falcons–Saints divisional tilt as two NFC South teams still searching for an identity in 2026. [1]

Why it matters

Front offices in the AFC North—Andrew Berry in Cleveland and Omar Khan in Pittsburgh—are wagering on contrasting 2026 timelines, with the Bengals and Ravens pressing from Cincinnati and Baltimore. Cleveland doubled down on rookie playmakers and special-teams range to tilt one-score games after moving Myles Garrett, while Pittsburgh bet that veteran QB stability plus DK Metcalf gravity would yield efficient throws even versus man coverage. Those choices showed up snap-to-snap against 73.3% man looks on Rodgers and a 56-yard kick greenlit by Cleveland. [2][6][7][10]

The gains and losses are concrete. Cleveland’s rookies produced 153 receiving yards out of 372 total, and Szmyt’s live leg changes fourth-down math in the opponent’s 38–43 band; but losing a force like Mason Graham while Garrett plays in Los Angeles invites leaks unless Berry buys reinforcements. For Pittsburgh, another week of tight man without a second separator burns clock on a narrow 2026 window despite Jaylen Warren’s 17 for 93 on the ground. Each missed layup throw into sticky coverage is a possession you can’t get back at age 42. [2][5][7][8][10]

Original analysis

Contrarian read

• Consensus: Cleveland’s Thursday win was a “Deshaun Watson resurgence” story.
• My read: It was roster construction and geometry. Berry drafted complementary rookies—Boston the boundary bully, Concepcion the YAC injector—and Todd Monken called spacing that created manufactured windows rather than hero balls; that combo accounted for 153 receiving yards, including Boston’s 60-yard shot that flipped field position before Szmyt cashed from 56 at 0:10. That’s repeatable structure, not a hot-hand mirage in Ohio. [2][3][4]

Back-of-envelope calculation

• Rookies’ game share: Cleveland produced 372 total net yards; Boston (4/89) + Concepcion (5/64) = 153 → 153 ÷ 372 ≈ 41.1% of total offense from two first-years. Show your work: 153/372 = 0.411. [2][3][4]

• Season pace: Through four games, Boston sits at 284 receiving yards; Concepcion at 144. Combined = 428; 17-game pace = 428 × 4.25 ≈ 1,819 yards, with Boston alone tracking ~1,207 (284 × 4.25). That’s WR1 production on a rookie deal if the clip holds through December 2026. [3][4]

2×2: How these teams create layups

• Axis 1 (Horizontal): Reliable separation vs man (High/Low).
• Axis 2 (Vertical): Trustworthy long-range kicking (Yes/No).
  • High separation + Long-range: 2018 Chiefs archetype—pass at will and still bank points.
  • High separation + No long-range: 2020 Bills early season—aggressive fourth downs.
  • Low separation + Long-range: Cleveland 2026—Monken manufactures space and Szmyt extends drives into points from the 38–43. [2][12]
  • Low separation + No long-range: Pittsburgh’s current trap—tight windows plus fewer bailout points, which strains Rodgers on every third down. [6][10][12]

Historical analogue (with year)

• The 2013 Ravens re-centered close-game math around Justin Tucker’s 61-yarder in Detroit, and John Harbaugh adjusted fourth-down thresholds accordingly that winter; Cleveland’s posture with Szmyt from 56 echoes that infrastructure shift. Monken doesn’t need Watson to be 2018 MVP-level if the kick map keeps adding 3s from the high-40s and low-50s late in halves. [2][12][13]

Named-stakeholder breakdown

• Andrew Berry (Browns GM): Trading Myles Garrett to the Rams in June 2026 forced a philosophical reboot; rookie WR hits plus Szmyt’s leg justify a tilt toward high-variance offense/special teams while the front recalibrates. Next move: a veteran DL rental who logged 300+ snaps in 2025 to soak base downs. [2][7][8]

• Todd Monken (Browns HC): Thursday showed a man-coverage answer key—horizontal stretches to Concepcion and verticals to Boston—while trusting long kicks to settle fringy drives in the 38–43. Standard fourth-down models support that shift in the middle of the field. [2][12]

• Omar Khan (Steelers GM): The WR room lists DK Metcalf and Michael Pittman Jr., yet 73.3% man looks say defenses aren’t scared by Roman Wilson or Germie Bernard right now. Unless one wins early in routes by Week 8, this tilts into a Jaylen Warren-centric offense out of necessity. [6][10]

• Aaron Rodgers (Steelers QB): The arm still punishes coverage busts, but window creation is the constraint. With a 21.6% pressure rate faced by Watson on the other sideline illustrating how man can throttle rhythm, Pittsburgh’s passing game needed layups that rarely materialized—and that’s roster/tactics, not birth certificate. [2][6]

2026 NFL Week 4 takeaways, properly framed

• Cleveland’s identity is no longer “waiting on QB salvation.” It’s rookies tilting the field, then special teams banking edges from 53–56 yards to survive a short-term Mason Graham absence and a post-Garrett front. That approach wins AFC North rock fights in October and December in Ohio and Pennsylvania. [2][7][8]

• Pittsburgh’s identity is misaligned with its quarterback. A 42-year-old needs easy throws; Thursday offered few, while Jaylen Warren produced 17 carries for 93 yards (5.5 YPC) and steadied the offense for the second straight week. The fix isn’t “more go balls to DK”; it’s forcing separation with motion/bunch and feeding RB/TE option routes until a second WR emerges. [5][10][11]

What others are missing

The Szmyt factor is strategic infrastructure, not a one-off highlight. Cleveland greenlit a 56-yard attempt with 0:10 left, which signals trust in the 38–43 yard-line band that changes Monken’s third-and-medium calls from low-percentage hero shots to schedule throws (drag to Concepcion, iso to Boston) and still bank 3 if stalled. In a 2026 AFC North that will likely swing on one-score scripts in Cleveland, Pittsburgh, Cincinnati, and Baltimore, that 3–5 yard shift in fourth-down thresholds compounds over 17 games. With Garrett gone and Graham nicked, Cleveland needs razor-thin special-teams edges, and Szmyt just posted proof-of-range in Week 4. [2][7][8][12]

What to watch next

1. By October 31, 2026, Denzel Boston will clear 500 receiving yards on the season (he’s at 284 through October 1), confirming a true WR1 trajectory. [4]

2. Over Weeks 5–8 (games ending by October 31, 2026), Jaylen Warren will average at least 18 touches per game and the Steelers’ neutral (1st–3rd, score within ±7) pass rate will fall below 51%, reflecting a run/TE-heavy correction. [11]

3. By the NFL trade deadline on November 3, 2026, Andrew Berry will add a veteran defensive lineman (DT/EDGE) who played 300+ snaps in 2025 to stabilize early downs while Mason Graham recovers. [7][8]

Sources

[1] Bleacher Report — “NFL Week 4: Winners, Losers, and Lessons” (Oct 1, 2026) — Week overview and Browns–Steelers framing; mentions rookies and Mason Graham.
[2] NFL Gamebook — Browns 27, Steelers 24 (Sep 30, 2026, Cleveland) — Official snap counts, scoring summary, total yards, and Szmyt’s 56-yarder at 0:10.
[3] ESPN Box Score — Steelers at Browns, Week 4 (2026) — Receiving lines for Denzel Boston (4/89) and KC Concepcion (5/64).
[4] Pro-Football-Reference — 2026 player game logs for Denzel Boston and KC Concepcion — Season-to-date yardage to project paces.
[5] Associated Press — Game recap (Sep 30, 2026) — Confirms Jaylen Warren’s 17 carries for 93 yards and key drive details.
[6] Next Gen Stats — Coverage/pressure splits for Week 4 (2026) — Man coverage rate (73.3%) vs Rodgers and pressure metrics.
[7] NFL.com News — Myles Garrett traded to Los Angeles Rams (June 2026) — Context for Cleveland’s defensive identity shift.
[8] Cleveland Browns — Week 4 injury report and postgame notes (Sep/Oct 2026) — Mason Graham status and rotational implications.
[9] Cleveland Browns — Todd Monken media availability transcript (late Sep 2026) — Philosophical hints on spacing and rookie usage.
[10] Pittsburgh Steelers — 2026 roster/WR depth (Metcalf, Pittman Jr., Wilson, Bernard) — Personnel that informs coverage responses.
[11] The Athletic — Steelers usage trends and neutral pass rate tracker (Sep–Oct 2026) — Baseline for expected run/pass adjustments.
[12] rbsdm.com — Fourth-Down Decision Model primer — Framework for how longer field goals shift optimal decision thresholds.
[13] ESPN — Ravens at Lions (Dec 16, 2013) recap — Justin Tucker’s 61-yard FG precedent for long-range strategic posture.

TL;DR

• Week 2’s quieter scoring hides obvious Week 3 waiver value: Adonai Mitchell’s 12 targets (33% share) are the sticky signal you pay for, not last week’s yardage outliers. [4], [5]
• Denzel Boston is not a gadget rookie—Cleveland already gave him a 95% route rate and a 16.5-yard aDOT; that combo scales fast in 12-team PPR leagues. [6], [7]
• Bryce Young’s hot start isn’t smoke; Carolina’s 34–3 win in Atlanta with 287 yards and three TDs feeds a real Week 3 test at Cleveland on September 27, 2026. [2], [9]

What the source said

Fantasy Life’s Week 3 “Waiver Wire Pickups & FAAB Advice” column argues that despite fewer touchdowns in Week 2 than Week 1, there are better underlying usage bets available, led by Bryce Young’s six TDs through two games and WRs who earned real roles, with concrete FAAB bands by league size. The piece highlights Cleveland in Week 3 for Young’s schedule context and frames bids around roster rates and opponent matchups. The thesis: ignore low-scoring noise; chase roles and usage that project forward. [1]

Why it matters

The real stakeholders are not “fantasy managers” in the abstract. They’re:

• 10–12 team redraft GMs who must decide whether to spend 12–18% FAAB on WR usage risers now or wait for the RB injury carousel that typically hits by Weeks 4–6.
• Guillotine-league players operating in a $1,000 FAAB ecosystem where survival, not ceiling, dictates spending cadence; a mispriced early splash can get you chopped before premium names hit the wire. [8]

In Week 3, those groups intersect with concrete NFL shifts: Adonai Mitchell’s elevation into a true Jets WR2 via a 12-target, 33% share game, Denzel Boston’s immediate downfield role in Cleveland (95% routes, 16.5 aDOT), and Bryce Young’s verified step forward after a 34–3 road win in Atlanta ahead of a September 27, 2026 trip to Cleveland. [5], [6], [2], [9]

Original analysis

Why this Waiver Wire Pickups & FAAB Advice For Week 3 hits different

Consensus: “Chase last week’s points.” My read: Chase the jobs that just solidified.

• Adonai Mitchell (Jets WR): Week 2 re-slotted New York’s room and put Mitchell on a full-time track opposite Garrett Wilson. He drew 12 targets (33.3% share), caught 6 for 63 vs. Green Bay, and played 80%+ snaps—bankable usage, not a fluke box score. [4], [5]
• Denzel Boston (Browns WR): Two games, seven catches, 154 yards, two TDs, with a 95% route rate and a 16.5-yard aDOT. That’s a ready-made vertical role, not a rotational rookie. Cleveland already schemed him downfield; that travels. [6], [7]
• Bryce Young (Panthers QB): 287 yards and three TDs in a 34–3 road win in Atlanta puts real signal on the board. He now walks into Cleveland in Week 3, the sort of environment that clarifies whether he’s a weekly streamer or back-end QB1. [2], [9]

Contrarian read

• The hot waiver list will hype Tre Tucker types off single-week yardage spikes. The smarter spend is on target and route monopolists—Mitchell and Boston—because volume compounds while splash plays regress. Mitchell’s 12 targets and 33% share are precisely the traits that persist; Boston’s 95% route rate plus deep usage means he only needs 6–8 looks to swing weeks. [5], [6]

Back-of-envelope math

• Mitchell projection: If his rest-of-season target share stabilizes at an attainable 22% (down from 33% in Week 2) on a 33-attempt Jets baseline, that’s 7.3 targets per game. In full PPR, at a modest 7.5 yards/target and 67% catch rate, you’re talking roughly 5.0 receptions (~10.0 PPR) and ~55 yards (5.5 points) per week before TD variance—call it ~12–13 PPR median, with spike potential as aDOT rises. Assumptions anchor to Week 2 usage. [5]
• Boston projection: Keep his 95% route rate, assume a 21% target share on a 32-attempt Browns baseline. That’s ~6.7 targets. With his 16.5 aDOT, a 55–60% catch rate yields 3.7–4.0 receptions and 60–70 yards (9–11 PPR) weekly, plus elevated TD odds because his targets are downfield and near the paint. Role > name here. [6], [7]

Named-stakeholder breakdown

• Joe Douglas (Jets GM): If Mitchell holds a 20%+ share by October 31, 2026, New York’s passing game shifts from one-outlet to two-outlet, and you get a weekly WR2/3 from the wire. That outcome validates early FAAB aggression. [5]
• Kevin Stefanski (Browns HC): Early usage says Stefanski is comfortable letting a rookie stretch vertically. If that continues into October, Boston’s per-target value dwarfs low-aDOT possession roles, and your bids should chase routes plus depth shots, not jersey seniority. [6], [7]
• Dan Morgan (Panthers GM): Young’s two-game line (six TD passes) underlines organizational stability around QB development; a credible showing in Cleveland turns a streamer into a medium-term hold in 10–12 team leagues. [1], [2]

2×2: Where to spend FAAB this week

• Axes: Usage (bankable vs. speculative) × Fantasy role (floor vs. ceiling)

Guillotine angle (where survival trumps upside)

• Public guillotine leagues often use $1,000 FAAB. Early-season survival means you prioritize stable weekly points over home-run profiles. Mitchell’s target-share profile fits survival bidding better than chasing last week’s yardage blips; Boston’s deep role is volatile week-to-week, but the route dominance keeps him viable if priced fairly. Think 4–7% on Mitchell and 3–5% on Boston unless your room is barren; keep powder dry for the first true “stud drop.” [8]

Historical analogue

• Amon-Ra St. Brown’s late-2021 breakout with Detroit (10+ targets in six straight from Week 13, 2021) showed how fast value compounds once a team condenses targets to two primary receivers. Mitchell’s Week 2 one-third share echoes that pathway—buy the consolidation before the box scores catch up.

What others are missing

The quiet catalyst behind Mitchell’s week-to-week viability is structural, not box-score: New York condensed routes and kept him on the field for 80%+ snaps while feeding him 12 targets (33.3% share). That pairing—full-time participation plus first-read volume—predicts stability in a way raw yardage never will, and it explains why a 6–63 line can be a green light rather than “just fine.” If you’re pricing FAAB as if he’s a flash in the pan, you’re ignoring the route and snap math that created his job. [4], [5]

What to watch next

1. By October 18, 2026 (Week 6), Adonai Mitchell will have at least three games with 8+ targets and two with 60+ receiving yards, locking him into weekly WR3 territory in PPR.
2. By October 11, 2026 (Week 5), Denzel Boston will record another 50+ yard touchdown or 80+ yard game, reflecting continued high aDOT usage in Cleveland.
3. By October 4, 2026 (Week 4), Bryce Young will rank top-12 in TD passes and passer rating despite a Week 3 trip to Cleveland, sustaining his early QB1/streamer case.

My take

If you have one premium bullet this week, fire it at Adonai Mitchell. Roles beat vibes, and his role just became sticky, tradable volume on a roster that needed a second grown-up target; spend 12–18% FAAB in standard 12-team PPR and lean high if you whiffed at WR in August. Add Denzel Boston where you start three WRs or reward big plays—he’s a cheaper path to wins than chasing last Sunday’s mirages. On QB, buy Bryce Young as a rest-of-season hold if he clears Cleveland with competence; if he does, stash him and stop streaming.

Sources

[1] Fantasy Life — Waiver Wire Pickups & FAAB Advice For Week 3 — Usage-first waiver logic, Bryce Young’s six-TD context, and practical FAAB bands.
[2] Panthers.com — Inside The Numbers: Panthers at Falcons, Week 2 (2026) — Confirms 34–3 final and Bryce Young’s 287 yards with three TDs.
[3] NFL.com — Game Center: Panthers at Falcons, Week 2 (2026) — Official scoreboard and team stat corroboration for Carolina’s blowout in Atlanta.
[4] CBS Sports — Jets vs. Packers Week 2 (2026) recap/box score — Lists Adonai Mitchell’s 6–63 receiving line and notes 12 targets.
[5] Team/usage tracker — Jets Week 2 snap counts and target shares — Documents Mitchell’s ~82% snaps and 33.3% target share.
[6] DraftKings Nation — Is Denzel Boston a must-add for Week 3? — Cites 95% route rate, ~21% target share, 2.4+ YPRR, and a 16.5-yard aDOT.
[7] ClevelandBrowns.com — Denzel Boston player page/notes — Verifies two-week totals (7–154–2) and role within Kevin Stefanski’s offense.
[8] FantasyPros — Guillotine League Strategy (2026) — Explains common

Snap’s $400M Bet on Perplexity: Why Snapchat Just Got a Lot More Curious

Snap’s announcement that Perplexity will pay $400 million to integrate its AI-powered search engine into Snapchat feels like one of those pivot moments you can almost hear in slow motion. The deal — a mix of cash and equity, rolling out early in 2026 — immediately lit a fuse under Snap’s stock and reframed the company’s AI ambitions from experiment to platform play. But beyond the market fireworks, this pact tells us something about the next phase of social apps: search and conversation are converging inside the apps people already use every day.

Quick snapshot

• Perplexity will be integrated directly into Snapchat’s Chat interface, surfacing verifiable, conversational answers to user questions.
• The $400 million payment is to Snap over one year (cash + equity) and revenue recognition is expected to start in 2026.
• Snap will keep its own My AI chatbot; Perplexity will act as an “answer engine” available inside chat, with Perplexity controlling the response content.
• The news came alongside stronger-than-expected Q3 results from Snap, and the stock jumped sharply on the announcement. (investor.snap.com)

Why this matters (and why investors cheered)

• Distribution = growth for AI startups. Perplexity gains nearly a billion monthly users as a built-in capability inside Snapchat — a shortcut to scale that usually takes years (and huge marketing). That distribution is worth a lot in today’s attention economy. (techcrunch.com)
• New revenue model for Snap. Instead of building and owning every AI layer, Snap is becoming a marketplace — a platform that offers high-quality third-party AI features and captures revenue for the placement. That’s a faster, less risky route to monetization than trying to train everything in-house. (investor.snap.com)
• User behavior is changing. People prefer getting answers where they already spend time. Embedding conversational search inside chat reduces friction and keeps attention and ad dollars inside Snapchat instead of sending users off to the open web. (reuters.com)

The practical trade-offs and questions

• Who controls the content? Snap says Perplexity will control its responses and that Perplexity won’t use those replies as ad inventory. That preserves a level of editorial and brand separation — but it also raises questions about moderation, factual accuracy, and how disputes will be handled when AI answers go wrong. (investor.snap.com)
• Data and privacy. Snap has claimed user messages sent to Perplexity won’t be used to train the model, but users will still have messages routed to an external engine. Transparency about data flows and safeguards will be crucial for trust — especially for younger users and privacy-conscious markets. (investor.snap.com)
• Economics vs. compute. Paying for AI placement is one thing; making the unit economics work long-term is another. Perplexity is effectively buying distribution today — but as usage scales, compute and moderation costs could balloon. Will revenue from the placement plus future monetization options offset those costs? Analysts flagged this as a watch item. (investing.com)

A competitive angle: Snap’s place among the AI arms race

Snap isn’t the only company stuffing AI into social. Meta, TikTok, X and others are all experimenting with conversational assistants, generative features, and AI-powered search. But Snap’s path is distinct:

• Platform-first, partner-driven. Rather than bake everything into a proprietary stack, Snap is inviting specialized AI companies into its app as first-class partners. That could accelerate innovation and let Snap remain nimble.
• Youthful audience, mobile-native context. Snapchat’s demographic — heavy on 13–34-year-olds — gives Perplexity a unique testbed for conversational search behaviors that other platforms may not replicate as cleanly. (investor.snap.com)

This approach could scale if Snap builds a robust ecosystem of AI partners (and if regulators or policy changes don’t intervene). Spiegel has signaled openness to further partnerships, hinting at a future in which different AI assistants sit alongside each other inside Snapchat for different tasks. (engadget.com)

Design and user experience implications

• Contextual answers inside chat feel natural: asking a quick question in a conversation or while viewing content is low friction and meets users where they already are.
• Verification and citations matter: Perplexity emphasizes “verifiable sources” and in-line citations. If executed well, that could distinguish Snapchat’s answers from hallucination-prone assistants and slow the growing distrust around AI outputs.
• Product sequencing is key: early 2026 rollout gives Snap time to AB test placements, UI patterns, moderation flows, and ad/product hooks — which will determine whether this is sticky utility or a novelty. (investor.snap.com)

Possible risks and blind spots

• Over-reliance on a single external provider. If Perplexity’s performance, reliability, or content decisions become problematic, Snapchat’s experience could suffer.
• Regulatory heat. As governments scrutinize algorithmic systems, an in-app AI that serves tailored answers to young users could draw policy attention on age protections, misinformation, or advertising rules.
• Cultural fit. Not all of Snap’s users will see value in an in-chat search engine. Adoption will depend on product framing, speed, trust signals, and how well the feature integrates into everyday use cases.

Snap’s playbook — what to watch next

• Product signals: how prominently Perplexity is surfaced, whether it’s opt-in, and how Snap handles user controls and transparency.
• Metrics: engagement lift, usage frequency per user, and whether this drives higher ad yields or subscription conversions for Snapchat+.
• Ecosystem moves: announcements of other AI partners or a developer program that lets more AI agents plug into Snapchat.

My take

This deal is smart theater and pragmatic strategy rolled into one. For Perplexity, access to Snapchat’s massive, young, mobile-native audience is a growth shortcut. For Snap, the pact buys relevance in the AI moment without assuming all the execution risk. The real test will be execution: whether conversational search becomes a daily habit inside chats or remains a flashy add-on.

If Snap gets the UX right (speed, clear sourcing, and easy context switching) and keeps control over moderation and privacy, it could redefine how a generation asks questions — not by opening a browser but by typing into the same chats where they plan their weekends, gawk at memes, and swap streaks. That feels like a small change with outsized ripple effects.

Final thoughts

Big-dollar partnerships like this one are shorthand for a larger shift: apps are turning into ecosystems of specialized AI services, and the companies that win will be the ones that make those services feel native, trustworthy, and undeniably useful. Snap’s $400 million deal with Perplexity is a bold step in that direction — one that could either cement Snapchat as a go-to AI distribution channel or become another expensive experiment if the execution falters.

Sources

• Snap Inc. — Snap and Perplexity Partner to Bring Conversational AI Search to Snapchat (press release). https://investor.snap.com/news/news-details/2025/Snap-and-Perplexity-Partner-to-Bring-Conversational-AI-Search-to-Snapchat/default.aspx. (investor.snap.com)

• Reuters — Snap tops revenue estimate on ad boost and unveils Perplexity AI deal, shares jump. https://www.reuters.com/business/snap-beats-revenue-estimates-ad-boost-unveils-deal-with-perplexity-ai-2025-11-05/. (reuters.com)

• Financial Times — Snap shares jump after $400mn deal with AI start-up Perplexity. https://www.ft.com/content/498f4128-59c4-4877-9889-dc6f3ab5e43f. (ft.com)

• TechCrunch — Perplexity to pay Snap $400M to power search in Snapchat. https://techcrunch.com/2025/11/06/perplexity-to-pay-snap-400m-to-power-search-in-snapchat/. (techcrunch.com)