AI Competition Radar: July 26–August 2, 2026

The July 26–August 2 window produced two usable result sets. In Addis Ababa, TENAx won the overall Cursor AI Hackathon Ethiopia with an Amharic health companion for people managing diabetes and hypertension. In medical imaging, MICCAI's Startup Village named CONVEA, SurgExplore, and NeuroPath as its first-, second-, and third-place startups. The next result dates are already visible for H0, UiPath AgentHack, FIND EVIL!, and OpenAI Build Week.

The Cursor AI Hackathon Ethiopia took place July 25–26 at Addis Ababa's Adwa Museum. The Ministry of Innovation and Technology said it was aimed at idea-stage AI startups and expected teams to turn a real problem into a minimum viable product during a 20-hour build. Cursor, Ethio telecom, Addis AI, ElevenLabs, Fal, Whisperflow, and local incubation partners were named in the organizer's event post. The promised support was practical rather than a published cash purse: tool credits, mentorship, technical help, and possible incubation with workspace, coaching, product validation, and investor or partner connections. 1

A TechAfrica News report said more than 300 developers, students, and innovators took part. A third-place team post gives the sharper breakdown: 300 developers across 60 teams. The event targeted healthcare, service delivery, and other local problems, but the public result record does not provide a complete official roster. 67

TENAx is the clear follow-up lead. The five-person team — Tesfamichael Abebe, Bemnet Mussa, Amanuel Merara, Abel Tesfa, and Gersum Asfaw — won first place in Healthcare & Lifestyle and was also described by a team member as the overall global winner. Its product is an AI health companion that speaks Amharic, helps people living with diabetes and hypertension manage their health daily, and gives clinicians a longer-term picture instead of a single short consultation snapshot. 2

The public post does not disclose a demo URL, customer, funding, or clinical partner. It does provide a direct founder-team contact path through Tesfamichael Abebe's LinkedIn profile. The first diligence request should be concrete: show the patient workflow, explain what is spoken versus text-based, and identify where the clinician receives longitudinal data. The competition result proves a strong build signal; it does not yet prove clinical efficacy or retention.

Pause Guard placed third overall. The five-person team — Yeabsera Lisanework, Kalab A. Bezabeh, Mikael Alehegn, Yostena Girma, and Sifen Woldemariam — says it built the product in 18 hours, its first hackathon. The team reports support from Addis Ababa University Startup Center and Addis AI, plus potential incubation with Ethio Telecom. Its public post does not explain what Pause Guard does, so the name should not be treated as a product description. 7

For a scout, the useful split is simple: TENAx has a legible problem and named team; Pause Guard has a placement and an incubation path but still needs a product explanation. The second-place project is not included here because a readable public project brief did not surface alongside the verified first- and third-place posts.

MICCAI announced its Startup Village podium on July 28. This was a result bulletin for a virtual pitch competition held on July 2, not a new event that concluded on July 28. Nine startups had been selected for the pitch; the format combined presentations, an expert-jury Q&A, and audience voting, with finalists advancing toward an in-person presentation at MICCAI 2026. 38

The three teams are not one comparable category beyond medical imaging. CONVEA is a registration and radiotherapy infrastructure play; SurgExplore sits inside a surgical workflow; NeuroPath is a diagnostic-biomarker pipeline. That distinction matters for outreach: a strong pitch result is a reason to request the technical deck, not a reason to assume the same sales cycle or regulatory burden.

MICCAI's winner announcement does not publish founder names, financing, customer traction, or product demos. The most useful public artifact beyond the announcement is the Startup Village recording. The result source does, however, give enough product specificity to make the first email different for each team rather than sending a generic "tell me more" note. 3

These entries are not confirmed winner lists. They are the next checkpoints where a named result may become available.

The first four checkpoints fall inside the next 7–14 days. Qwen Cloud is just beyond that immediate window, but its five-track structure and $70,000 purse make it worth keeping in the wider pipeline. None of these pages should be treated as evidence that a team won until the result page names the team and the project.

If the goal is a short follow-up list rather than a long reading list, start with four different questions:

Pause Guard belongs in a second pass until its product description is public. The same rule applies to H0, UiPath AgentHack, FIND EVIL!, and OpenAI Build Week: wait for named winners and inspect the artifact before treating a scheduled announcement as a sourcing signal.

The public record for this week is uneven, but the usable leads are specific. TENAx offers the clearest founder-and-problem match; the MICCAI podium offers three technical theses that can be compared on validation and deployment; the next batch is a dated queue, not yet a winner list.