Parag Agrawal

We’re thrilled to highlight our new collaboration with @harvey, the leading AI platform for legal and professional services. Harvey uses Parallel’s accurate, relevant, and fresh web search across their platform to retrieve valuable public context for their legal AI workflows. Together, we’re helping Harvey expand its best-in-class AI legal platform to over 60+ countries by collaborating on a specialized index of hard-to-reach international legal domains, built on our custom web search infrastructure. To learn more about Harvey and Parallel, read our blog: parallel.ai/blog/case-stud…

@m_atoms you should check out @p0 monitor API to build a hyper-local feed for this. may help content liquidity

Adverse media is a major KYB bottleneck but banks are wary of automation. Our new Taktile Labs + @p0 benchmark shows AI agents can now exceed human performance on this task. We partnered with Parallel to test an adverse media agent built on top of 7 frontier AI models with 47 real businesses. tldr: High performance of AI in this task signals the level of automation we are likely to see in core compliance work. our top findings: 1. AI agents produce better quality evidence than humans. This is based on a 20-point rubric covering evidence quality, ability to identify the correct entity, and risk assessment of findings. Human analysts average 13.50 while AI agents score 14.60. 2. A hybrid agent x human setup reduces workload by 93% Full automation is not necessary for high impact. A hybrid model where an agent screens first and a human reviews uncertain cases reduces analyst workload by 93%. 3. Even the “cheaper” models are still very good at fulfilling the task. Agent configurations using some of the less expensive models can still match human performance. Costs will continue to go down as new models are released and accuracy will keep getting better. 4. How you configure the agent matters more than the model you choose. The investigation strategy and prompt you put behind the agent has a much bigger impact on results than the model itself. 5. Agents do not hallucinate in this task but they can “over-flag”. This is huge. When we see false positives from agents, they are not fabricated sources (0% hallucination rate in this dataset). However the agent will sometimes flag sources too eagerly. Huge thanks to David Ahn and @maximilianeber for leading the research and @paraga and Sahith Jagarlamudi for your partnership on this. See the full study: taktilelabs.ai/benchmarks/kyb…

Gumloop's agents now use Parallel Monitors as a trigger for downstream work. Just tell your agents to watch for signals across the open web and let them do the rest. Learn more about Monitor: parallel.ai/products/monit…

This week, we officially welcome Liam Mulcahy and Graham Moreno to lead and scale GTM at Parallel. Liam joins us from @kleinerperkins, where he helped kick-start GTM operations in early-stage companies like @tryprofound, @harvey, @glean, @windsurf, and most recently, Parallel. Graham previously led world-class sales organizations at @windsurf /@cognition, @grafana, and @MongoDB, where he and Liam were colleagues. Want to work with them? Join us parallel.ai/careers

x.com/i/article/2034…

Time to reveal who let the 🦞 out ;) Today, @Taktile launches Taktile Labs. We dropped the lobster on Wall Street to ask the question: are banks ready for autonomous agents? With our applied AI research institute, we aim to bridge the gap between what frontier models can now do - and what regulated institutions need in order to trust it. Our first benchmark shows the latest models can beat human accuracy on very complex banking tasks: 96%+ vs. 89% in financial spreading. The models are ready. Now the industry needs evidence, benchmarks, and practical frameworks to ensure they work reliably at scale. That is what Taktile Labs is built for. AI is coming to financial services - let's make sure we can trust it. Excited to drive this with a stacked internal team and many incredible individuals on our Research Council and Advisory Board. Thanks to Bradesco’s Fagner Abreu, Parallel’s @paraga , Founder, Investor, and Morgan Stanley Lead Director Tom Glocer, Harvard Business School Professors Robin Greenwood and Karim Lakhani, Harvey’s Ben Liebald, Camunda’s Daniel Meyer, Cursor’s Jonas Nelle, ROC Partners’ Tina Reich, Equifax’s Harald Schneider, Suno’s @MikeyShulman, Intuit’s Henry Venturelli, Allianz Partners’ Pieter Viljoen, Flexcar’s Michael Zambrano, and Varo Bank’s Jill Zucker Sheckman. Learn more at: taktilelabs.ai (nothing AI generated about it btw, we worked with NYC artist @AndrewLoganAMW to build the lobster from scratch)

AI answer engines are the new search, and getting cited by ChatGPT, Gemini AI overviews, and Copilot is the future of discovery for brands. @tryprofound built the marketing platform for the AI era to win this shift, with tools powered by Parallel for brands to create high-quality content that gets cited by AI.

Parag Agrawal (@paraga) is speaking at Compute Conference. Founder of @p0. Building search infrastructure for AI agents. $100M Series A, $740M valuation. Previously CTO and CEO of Twitter. Join us March 8–9 at Chase Center, SF. Tickets: go.daytona.io/9DmGPuN

Mozilla built @tabstack to let AI agents browse the web the way humans do: by clicking, scrolling, form-filling, and extracting page information. They use Parallel’s Search API to handle what comes before all of that.

We just shipped a brand new Company Enrichment cookbook with @supabase: supabase.com/partners/integ… It uses Supabase Edge Functions to make database-native calls to Parallel’s Task API, enabling web-based enrichment directly in your tables. Demo and cookbook in thread: