01 / Build This Weekend
Sidewalk delivery robots get stuck because their path planner can't keep up. A new convex optimization method just made it 10x faster.
What just became possible
A November 2021 arXiv paper introduced superlinearly convergent trajectory planning for autonomous systems with complex collision avoidance constraints. Onboard, no trust-region limitations, no offline pre-computation. Subsequent work in 2024-2025 brought it to real-time embedded deployment. The math problem that has frustrated robotics teams for two decades is now solved engineering.
Why now
Last-mile delivery robot market: $1.5B in 2026 to $11.5B by 2035 at 24.5% CAGR. Delivery robots overall: $686M in 2026 to $7.62B by 2034. Starship, Kiwibot, and 30+ competitors are deploying at scale. Stuck states are the #1 operational cost driver and the bottleneck on city expansion permits. 172,300 robotics engineering roles in the US by 2029.
What you'd build
A real-time motion planning SDK that guarantees collision-free trajectories without pre-computation. License to last-mile delivery robot startups, warehouse automation companies, and industrial robotics. Pricing: per-robot license plus support. Average robotics engineer: $142k; senior specialized: $180k+.
Who's already moving
Realtime Robotics (RapidPlan, hardware-accelerated). Covariant (Brain platform). Boost Robotics (YC, data center manipulation). NVIDIA Isaac Sim, Newton 1.0 physics. Amazon Robotics (in-house). Starship Technologies. The whitespace: a generic SDK that drops into ROS 2 and works across robot brands, with provable guarantees on dynamic obstacle handling.
The gap
Guarantees that hold in cluttered, high-pedestrian-traffic environments. Edge-case handling without manual tuning. Amazon Research Awards ($50k cash + $50k AWS credits). NVIDIA Academic Grants. ARM Institute (manufacturing motion planning). EU H2020. The deal: open-source core, paid commercial wheel with SLA on stuck-state reduction.
02 / AI Makes This Possible
OpenAI moderation costs $10-$50 per million tokens. Distilled custom filters run 5-100x faster, without labeled data.
What just became possible
A January 2026 arXiv paper showed automated training of custom, policy-aligned moderation filters using synthetic data from foundation models. No human labels. Runs 5-100x faster than reasoning models. Policy-as-prompt plus distillation plus synthetic data closes the loop. Cultural, dietary, age-specific rules all expressible without retraining.
Why now
Content moderation services market: $13.94B in 2026 to $42.36B by 2035 at 13% CAGR. Automated moderation segment: $1.48B in 2026. EU DSA, UK Online Safety Act, US TAKE IT DOWN Act all in force. Discord-likes, dating apps, UGC forums managing 100k+ daily posts are the immediate buyers. Trust and Safety Analyst salary: $73k/year, but the API spend is what scales.
What you'd build
A developer API that generates custom moderation filters in seconds. Buyer types in their policy, system synthesizes training data, distills a fast classifier, ships an endpoint. Sell to Heads of Trust and Safety at mid-to-large social platforms. Replace $10-$50/M-token LLM API spend with $0.50/M-token specialist endpoints.
Who's already moving
Moonbounce ($12M raise in April 2026, 40M daily reviews, 100M DAUs). Stream (AI Moderation API, cultural fluency focus). Estha (no-code moderation, customizable severity). Hive Moderation (94% accuracy on AI image detection). Plus the LLM providers themselves (OpenAI, Anthropic) bundled into their APIs. The whitespace: distillation-first, policy-as-prompt, sub-second SLA.
The gap
Liability framework for automated moderation decisions. Multi-jurisdiction policy compliance. Languages beyond English-Spanish-Portuguese. Alibaba Cloud AI Catalyst ($500k-$3M). Cohere Labs Catalyst Grants. NSF NAIRR (up to $35M for ops centers). The land grab: get to 100M DAU customer count, then defend on policy quality.
03 / Deep Tech Bet
A single ivermectin shot could kill the mosquitoes that bite anyone in a village for two months. The veterinary regulatory pathway is the moat.
What just became possible
A July 2025 bioRxiv paper validated MedinCell's mdc-STM-001, a single-injection long-acting ivermectin formulation that holds mosquitocidal blood concentrations for over 60 days. The BEPO sustained-release technology hits WHO criteria for vector control. Veterinary formulation creates a community-wide mosquito-killing barrier when livestock are treated.
Why now
Annual global malaria control spend: $4.3B on IRS, LLINs, SMC. Insecticide and drug resistance is rising. Outdoor biting and zoophagic vectors are not addressed by bed nets. MedinCell has $3M from Gates Foundation for the program. The IMPACT consortium (Unitaid-funded) advances the human-and-veterinary crossover. First-in-human trials are planned.
What you'd build
A long-acting ivermectin veterinary formulation for cattle, applied at community scale in malaria-endemic regions. Sell to NMCPs and Global Fund implementing partners. Or co-develop the human formulation through MedinCell's pipeline. The regulatory complexity (veterinary plus human public health) is the moat against generic pharma.
Who's already moving
MedinCell (mdc-STM, lead candidate, 3-month subcutaneous). IMPACT Consortium (Unitaid, 6-month formulation). ISGlobal/Barcelona Institute for Global Health (regulatory pathway). Medicines for Malaria Venture, Target Malaria, Malaria Consortium on adjacent work. No second commercial pipeline besides MedinCell's BEPO platform.
The gap
FDA 505(b)(2) for human use is 5-7 years and $50-100M. Veterinary pathway plus environmental risk assessment for non-target organisms. Gates Foundation GHIT ($5M/project). Unitaid Innovative Vector Control ($10M/project). IVCC Next-Gen tools. The first commercial cattle-deployed formulation with verified community-level malaria reduction wins the next decade of vector control procurement.
Pyrethroid-resistance mechanisms vary by district. Clothianidin and chlorfenapyr break different mechanisms. The right mix saves three years of failed control.
What just became possible
A January 2024 bioRxiv paper identified resistance markers and validated clothianidin and chlorfenapyr efficacy in pyrethroid-resistant Anopheles gambiae in coastal Cote d'Ivoire. Clothianidin works on oxidase-dominant resistance. Chlorfenapyr works on metabolic resistance. Choosing the right one per district turns broken control programs into working ones.
Why now
Chlorfenapyr market: $0.5B in 2026 to $0.8B by 2035 at 5.8% CAGR. Broader insecticides driven by public health: $40.68B by 2030. The New Nets Project averted 13M malaria cases and 24,600 deaths annually with dual-AI nets. UNICEF and the Global Fund are the procurement bottleneck. WHO recommends pyrethroid-chlorfenapyr nets strongly for resistance areas.
What you'd build
Not a new net or IRS formulation. Build the resistance-marker testing service and the district-level recommendation engine that tells NMCPs which intervention (Interceptor G2 vs Royal Guard vs SumiShield IRS vs Fludora Fusion) fits their resistance profile. Sell as a procurement support contract.
Who's already moving
BASF (Interceptor G2 chlorfenapyr ITN). Bayer (Fludora Fusion clothianidin IRS). Sumitomo Chemical (SumiShield clothianidin IRS). IVCC coordinates. PRONet Duo (chlorfenapyr + bifenthrin) in WHO prequalification. No commercial resistance-mapping plus recommendation service. Academic surveillance work exists but isn't sold to procurement.
The gap
WHO endorsement of resistance-driven procurement decisions. Field-deployable resistance assays at scale. Real-time data into procurement workflows. Gates Foundation, Unitaid, IVCC, GiveWell ($41M to AMF for DRC) fund the surrounding ecosystem. The deal: anchor with one NMCP, expand to the regional procurement bloc, become the data layer for the Global Fund Revolving Facility.
05 / Watch This Space
Mpox Clade I and chickenpox look identical to the eye. A rapid lateral flow test could replace seven days of waiting for a PCR result.
What just became possible
An April 2025 medRxiv paper validated lateral flow assays differentiating Clade I mpox from varicella zoster virus with 96% sensitivity and 100% specificity. The SIGA Technologies license from Vanderbilt covers the specific monoclonal antibodies. The WHO Emergency Use Listing process for mpox diagnostics is open and active.
Why now
19 African countries reported 40,000+ suspected mpox cases in 2024, most unconfirmed due to testing capacity. WHO and Africa CDC declared public health emergencies in August 2024. $150M+ mobilized globally. Ministries of Health spend on sample shipping (3-7 day delays), expensive RT-PCR cold chain, and clinical-diagnosis misallocation of vaccines. Mpox diagnostics market: $2.06B by 2026.
What you'd build
A low-cost rapid lateral flow diagnostic kit that differentiates Clade I mpox from VZV in minutes. WHO EUL pathway. Sell to Ministries of Health and National Public Health Institutes in DRC, Nigeria, CAR, Burundi. Pricing: under $5/test, with local manufacturing partnerships through Africa CDC's PAMTA initiative.
Who's already moving
JOYSBIO (general mpox rapid test, ISO 13485, 3M kits/day). Contipharma (LAMP + RDT, DRC authorization, WHO EUL evaluation). Penn State (plasmonic prototype). Abbott Alinity m, Roche cobas, Cepheid Xpert Mpox (all PCR, lab-based). SIGA Technologies (poxvirus mAb license from Vanderbilt). The whitespace: differentiated lateral flow that solves the misdiagnosis problem at point of care.
The gap
Sourcing specific monoclonals at scale and low cost. WHO EUL approval during an active outbreak (bureaucratic but feasible). Local manufacturing in Africa via the EU4Health PAMTA Initiative (€9.4M). Unitaid Integrated Health Diagnostics call. Africa CDC plus FIND capacity building. First to ship 1M validated kits to DRC wins the next outbreak cycle.
See you next week.
- Theis