AI Competition Radar: June 8–15, 2026

Scout signals this week

This is a build week, not a results week. Four active hackathons are still mid-flight, two large events wrapped but haven't published winners, and YC S2026's first public launches are trickling onto the Launch YC page with real traction numbers attached. If you're scanning for outreach targets today, the highest-density surface area is those YC launches — five teams went public in this window, with founder names and upvote counts to prioritize the queue.

The structural story is participation scale. Google Cloud Rapid Agent closed at 14,526 participants, the largest AI hackathon on record for a Devpost-hosted event this year. 1 Band of Agents crossed 4,000 participants at the halfway mark. 2 USAII Global AI Hackathon kicked off its build week with 5,855 registrants, of whom 804 teams qualified through a prerequisite AI Readiness Qualifier. 3 Three separate events, three different crowds, none of them with published winners yet.

The one area with confirmed names: Peachscore Cohort 31 announced 74 new startups on June 9, and YC S2026 founders began going public this week. Both offer direct contact surface right now.

Hackathons: in progress

Band of Agents Hackathon (lablab.ai + Band.ai)

Three tracks are running: Regulated & High-Stakes Workflows (16 submissions), Internal Enterprise Workflows (15), and Multi-Agent Software Development (13). All projects must coordinate a minimum of three agents using Band.ai's shared communication layer.

The community vote leaderboard at day four: 2

Rank	Team	Project	Votes
1	EquiSaaS BD	EquiPulse AI	15
2	AI FOR LABLAB_AI	Aether Labs: Multi-Agent E-Commerce Orchestration	4
3	Yousun Apexia	YOUSUN Secura	2
4–9	Various	—	1 each
10	—	OpenMind Nexus	0

EquiPulse AI stands out as the community leader. Built by EquiSaaS BD, it's an offline-first business intelligence platform aimed at merchants in emerging markets — using Gemini 1.5 Flash, DuckDB, local Llama models, multimodal OCR trained on Bangla invoices, and hybrid Graph RAG for seasonal demand forecasting. 4 The thesis is straightforward: data infrastructure for markets where cloud connectivity is unreliable and bookkeeping runs on paper. If the team converts the hackathon into a pilot, the natural next step is a low-cost SaaS tier designed for sub-$10/month pricing in South Asia.

Aether Labs (by AI FOR LABLAB_AI) takes a different angle — a multi-agent e-commerce orchestration platform using OpenAI, LangChain, n8n, and Supabase, with an Agent Decision Timeline feature for auditability. 5 Explainability for multi-agent decisions is a real gap in the current tooling landscape; a standalone audit layer could be worth more than the orchestration itself.

YOUSUN Secura (by Yousun Apexia) is a multi-agent governance command center that intercepts sensitive AI actions — policy violations, security exceptions, compliance flags — before execution, then routes them for human review. 6 The positioning sits squarely between "AI compliance" and "agentic guardrails" — a category getting regulatory attention in financial services.

Other notable entries across all tracks: Recourse (five AI agents adjudicating disputed insurance claims, with a tamper-evident audit trail), HireGuard (four agents auditing hiring packets for EEOC and pay-transparency compliance), SafeHands AI (fraud detection via multi-modal computer vision for logistics and insurance), and TELMED AI DOCTOR (medical diagnosis by text, voice, and image for underserved African communities using Gemini 3 Flash and Claude). 7

Band of Agents Hackathon — community voting leaderboard at day 4 of 8 — Band of Agents Hackathon (lablab.ai) — 4,006 participants, 3 tracks, $10K+ prize pool 2

Scout note: Community votes are not official results — formal judging happens after the June 19 close. But 15 votes versus 4 votes for second place suggests EquiPulse AI has differentiated itself clearly enough that community reviewers noticed. Search "EquiSaaS BD" on LinkedIn before the official announcement.

Google Cloud Rapid Agent Hackathon — closed, winners July 13

Organizer: Google Cloud + Devpost | Closed: June 11 | Participants: 14,526 (up from 13,697 the prior week) | Prize pool: $60,000 across 6 partner tracks | Winner announcement: July 13, 2026 1

The six partner tracks and their prize buckets: Arize, Elastic, Fivetran, GitLab, MongoDB, and Dynatrace — each awarding $5,000 / $3,000 / $2,000 for first through third. All projects required Gemini via Google Cloud Agent Builder plus integration with at least one partner's MCP server.

No shortlist or finalist gallery has been published. The Devpost page shows "Winners announced soon" with July 13 confirmed. Judging begins after submission close, which means the first visible signal — a project gallery tab with ranked entries — may not surface until mid-July. Mark July 13 on the calendar, not earlier.

USAII Global AI Hackathon 2026 — build week active

Organizer: USAII | Format: Virtual, with awards June 27 | Build week: June 14–21 | Participants: 5,855 registered; 804 teams qualified | Prize pool: $15,000 cash plus USAII AI Certification scholarships 3

The build week kicked off June 14 via livestream. Three student tracks: High School (Grades 9–12), Undergraduate, and Graduate (Master's and PhD). Each track carries its own challenge theme — community support and climate action at the high school level, productivity and public services at undergrad, human safety and public policy at the graduate level.

A two-phase qualifier separated the field: 5,855 Devpost registrants were down-selected through an AI Readiness Qualifier (June 7–10), with 804 teams advancing to the build phase. Judging runs June 22–25, awards ceremony June 27. Devpost profiles for all 804 qualified teams are visible now — a searchable pipeline of student builders across all three education levels.

Other hackathons: pending results

The AI Hack (Devpost, 86 participants — up from 78 last week): "Winners announced soon," with next milestone June 18. Student-only, 5 non-cash prizes. 8
AI Season of Code (36 participants): Submission deadline was today (June 15 at 6:00am EDT). Student-only, 5 non-cash prizes. 9

Pitch competitions and accelerators

Peachscore Cohort 31: 74 startups, 15 countries

Peachscore's 12-month equity-free accelerator has now supported over 1,500 startups from 100+ countries; portfolio companies have collectively raised over $218M. 10

The industry breakdown for Cohort 31: Information Technology (14 startups), Software (13), Healthcare (12), Education (5), Biotechnology (2), Blockchain (2), with the remainder spread across food and beverage, hardware, energy, real estate, legal, media, retail, supply chain, sustainability, travel, and security.

Selected names from the cohort: Dakhila AI (education), DearMoney (financial literacy), CaretEDA (hardware and EDA tools), Tavisha Robotics (healthcare robotics), QuantZero (quantitative finance), SecureHive (AI security), Claimb.ai (insurance claims automation). 10

Peachscore founder Peyman Shahmirzadi noted that Cohort 31 "reflects the global nature of modern entrepreneurship and the growing variety of opportunities being pursued by early-stage founders worldwide." 10

Scout note: 88% Pre-Seed is the right stage for first contact. Peachscore provides a public startup directory; all 74 teams have profiles. Given the equity-free structure and global composition, these teams are frequently available for calls without a warm intro. Insurance (Claimb.ai), healthcare robotics (Tavisha), and quantitative finance (QuantZero) are three sub-verticals with recent investor demand signal.

YC S2026 launches: Ontora leads with 110 upvotes

Five S2026 batch companies posted on Launch YC during the June 8–15 window. This is the clearest scout pipeline the week produced — team names, founder handles, upvote counts, and project one-liners are all public. 11

Company	Launched	Upvotes	One-liner	Founder
Ontora	June 8	110	AI-powered discovery layer for enterprise AI transformation	David Korn
Result	June 10	16	Platform for entrepreneurs to start and operate their businesses	Aaryan Kushwah
Lattice Health	June 10	11	Monitoring and governance for deployed medical imaging AI	Christine Park
Akkari	June 9	8	Autonomous customer ops — agents that close the loop with every prospect	Jeffrey Byun
Aster	June 13–14	24	Autonomous research lab orchestrating thousands of agents in parallel	Emmett Bicker

Ontora is the week's strongest signal by upvote margin. David Korn frames the product around a diagnosis most enterprise buyers already feel: "Every company we talk to has the same story: they bought Claude licenses for their teams but merely use it to rewrite emails. The AI transformation stalled because nobody mapped what to actually automate." 12 Five enterprise design partners, 150+ inbound demo requests, and $700K raised pre-fundraising suggests the messaging is converting. The 80+ inbound VC calls he reported are a ceiling-pressure signal — if you haven't reached out yet, the window for first contact without a crowd is closing.

Aster (founded by Emmett Bicker, formerly of Magic RL) billed itself as "the first YC Neolab" — an autonomous research lab that runs thousands of agents in parallel. 13 The team claims a world record on the ProteinGym benchmark: completed in 30 minutes, 57× faster and at one-third the cost of prior approaches. 13 Bicker: "Aster isn't a single agent. It is the first autonomous lab orchestrating thousands of research agents in parallel." 13 The ProteinGym benchmark is a standard evaluation for protein fitness prediction — a hard computational biology task, not a synthetic test, so a 57× efficiency claim is checkable and specific.

Lattice Health (Christine Park) is in a narrow but high-value gap: governance and monitoring for medical imaging AI that's already deployed. Regulators now require post-market surveillance for AI/ML-enabled medical devices; this is the compliance infrastructure play, not a care-delivery play.

The YC company directory lists several additional S2026 AI companies that recently made their profiles public: Mireye (geospatial AI API for agents, San Francisco), Glen (organizational knowledge agents, San Francisco), RealPact (real estate AI agents for brokerages, San Francisco), GUILD (AI defense contractor automation, New York), and Touchmark (AI infrastructure, San Francisco). 14 The full S2026 batch list has not been published on ycombinator.com as of June 15 — individual companies are surfacing through the launches page and directory, not a single batch announcement.

Google Cloud Rapid Agent Hackathon banner — 14,526 participants, winners July 13 — Google Cloud Rapid Agent Hackathon — largest AI hackathon closed this cycle; $60K across 6 partner tracks 1

Pending pipeline: chronic gaps

Two competitions remain in judging with no published results — a situation consistent for the third consecutive week.

Web Data UNLOCKED (Bright Data + lablab.ai): Ended May 31 with 2,310 participants, 767 teams, and 294 final submissions across three tracks (Finance & Market Intelligence, GTM Intelligence, Security & Compliance). Status as of June 15: "Judging in progress" — 15 days after close, no winner list published. Community votes total 896, with AI-driven real-time web intelligence by Dominify leading at 89 votes, followed by Auto VIN Scout by VIN Scout AI (73 votes). 15 The judging message states results will publish as soon as review wraps.

Apart Research SPS (Secure Program Synthesis Hackathon): Ended May 24 with 328 sign-ups and 61 entries. Status as of June 15: no winner list published, 22 days after close. The event page states only "This event has concluded." 16 Top teams were offered fast-track consideration for the SPS Fellowship (June–October 2026, mentors include Erik Meijer and Shriram Krishnamurthi). Apart Research's public communications since May 26 have shifted to promoting the Global South AI Safety Hackathon on June 19–21.

Both competitions will be tracked until results publish. Given the fellowship pathway in SPS, the signal when it arrives will be worth the wait.

Scouts' calendar: June 19–29

This weekend — highest event density of June

Three events land in the June 19–21 window:

Global South AI Safety Hackathon (Apart Research + Schmidt Sciences) | June 19–21 | Hybrid — 8 local hubs across Latin America, Africa, and Asia | $6,000 in prizes distributed by region: Latin America $3,000 (3 winning teams), Asia $2,000 (2 teams), Africa $1,000 (1 team) 17

Sign-ups surged from 328 to 734 between the prior issue and this week — a 124% increase. 17 Hub cities: Buenos Aires, São Paulo, Bogotá, Mérida, Guadalajara (Latin America); Cape Town (Africa); Bengaluru, Hanoi, Ho Chi Minh City, New Delhi (Asia). Four tracks: Technical AI Safety, Governance/Policy, and locally-tailored problems per region. Best teams move to the Apart Fellowship for continued research and mentorship.

For scouts tracking AI safety research talent outside the US/UK cluster, this is the highest-density opportunity on the June calendar. The fellowship pathway makes this a multi-stage funnel, not a one-off competition.

CalHacks AI Hackathon 2026 | June 20–21 | In-person, UC Berkeley MLK Building (2495 Bancroft Way) | $100,000 in prizes | 1,300+ expected participants, 300+ projects anticipated 18

Applications closed May 17 — this is not open for new registrations. If you're confirmed: four tracks (Ddoski's World, Ddoski's Toolbox, Ddoski's Lab, Ddoski's Playground), 24-hour format, opening ceremony June 20 ~10 AM, closing June 21 ~6 PM. Sponsors include Anthropic, Midjourney, Annapurna Labs (Amazon Web Services), Fetch AI, Redis, and Adobe. With $100K in prizes and Berkeley's CS network, this typically produces teams that convert hackathon builds into seed-round pitches within 90 days. The project gallery on devpost will be the search surface after the event closes.

Microsoft Build AI Hackathon (India) — finalists announced June 19. 19 India-only, professionals with 3+ years of experience. Official platform details remain limited to social posts rather than a Microsoft landing page; treat as low-confidence until confirmed via Unstop or official Microsoft India channels.

Coming up — flag now, act this week

Event	Dates	Format	Prize	Deadline to apply	Key detail
Qualcomm × Meta ExecuTorch Hackathon	June 27–28	On-site, San Francisco (625 2nd St)	Meta Quest 3 512GB + Ray-Ban Meta AI Glasses per winner	Today, June 15	30-team cap; proposal-based; build on-device AI with Samsung Galaxy S25 Ultra hardware; judged on Technical Implementation (40%) and Application Use Case & Innovation (25%) 20
USAII Global AI Awards	June 27	Virtual (livestream)	—	—	804 qualified teams from the current build week; awards ceremony 10:00 AM ET 3
Google Cloud Rapid Agent winners	July 13	Devpost announcement	$60,000 total	—	14,526-participant pool; track winners across Arize, Elastic, Fivetran, GitLab, MongoDB, Dynatrace 1

Farther out — open applications

Event	Dates	Format	Prize	Status
RAISE Summit Hackathon	July 4–9	Paris + Online	Up to $150,000	Applications open 21
AMD Developer Hackathon Act II	July 6–11	Online	$10,000	Applications open; $100 AMD Developer Cloud credits per participant 21
Band of Agents winners	~June 19–22	—	$10,000+	Current hackathon closes June 19; expect results 2–5 days after

Coverage window: June 8–15, 2026 (UTC-8). Founder contact information is included only where publicly disclosed. Web Data UNLOCKED and Apart Research SPS results remain outstanding; the next issue will report those announcements when published.