←
Lena Hall11,318 posts
Lena Hall avatar

Lena Hall

@lenadroid

Senior Director, Developers & AI • Founder of Droid AI • Forbes Tech Council • LLMs • Data • AI/ML • Ex AWS, MSFT Research • Opinions are mine

◷ Joined December 2012▧ 1,094 Photos and videos
Lena Hall avatar
Lena Hall@lenadroid·11 Jan 2016

New generation of ladies who are gossiping about distributed systems, FP, pragmatic architecture, chaos engineering and fast data analytics.

Lena Hall avatar
Lena Hall@lenadroid·2h

your AI doesn't need to hallucinate to be wrong, bad data is enough and the more decisions we delegate to AI, the more expensive those mistakes can become imagine a retailer using an AI agent to decide where to open its next store the agent can analyze demographics, compare locations, and make recommendations but what if the location data is outdated? or customer and property records don't match across systems? the agent could reason perfectly and still recommend the wrong location and before you can fix that, your team might spend weeks just figuring out how to get the right data into the system that's what caught my attention about @PreciselyData AI Studio it's a free catalog of working apps, agents, and skills that you can run, download, and customize. They're built on Precisely’s capabilities, including data quality, integration, governance, enrichment, and location intelligence so instead of spending all your time figuring out how to connect and prepare enterprise data, you can start with a working example and adapt it to your business you can also use the models and tools your team already has I like this approach because the goal isn't to build another impressive AI demo. it's to build something that can make a useful decision with information the business can trust explore Precisely AI Studio: fandf.co/3X0Ov1m #PreciselyPartner

Post attachment
Lena Hall avatar
Lena Hall@lenadroid·Nov 12

Say what you want, but energy is what makes magic. When you are in the right, most authentic energy state, the wildest coincidences just happen, in the best way.

Lena Hall avatar
Lena Hall@lenadroid·Nov 6

“hey, model, can you just make yourself better at this one thing?” “sure, i just rearchitected my weights. all done” fascinating

1/ What if an LLM could not just edit a model's weights, but re-architect it as it trains? We tried just that. The resulting architectures exit the known territory of transformers: the coding agents spliced code directly into the network. On nanochat, the loss fell below ordinary training after each new code layer was introduced. On algorithmic reasoning, accuracy rose from 55% to 84%. We even let DeepSeek-V4.1-Flash act as the agent modifying itself, raising its AIME and HMMT 2026 score from 64% to 71%. To do all this, the agents built a set of tools we’ve never seen before, which we call mechanistic intervention! Blog: doubleai.s.gy/llm-redesign

Post attachment
Lena Hall avatar
Lena Hall@lenadroid·Nov 4

back to SF next week! looking forward to AI Summit, AI Engineer Code Summit, and a couple of private events say hi if you’re there ✨

Post attachment
Lena Hall avatar
Lena Hall@lenadroid·Nov 3

if you’re good at building software, designing, teaching, or solving one specific problem, you probably already have ideas that could be businesses turning them into one means research, offer, website, finding people, and outreach. that stuff usually stays disconnected if you’re doing it in pieces i made a new video where i take one of my existing projects and use @pazidotai to run the whole thing end to end it gives you an operating officer that coordinates specialist agents under one business context. they handle the research, product work, and outreach together. you review and decide what moves forward try it here: links.pazi.ai/lena

1:46
Lena Hall avatar
Lena Hall@lenadroid·Oct 31

don’t let me anywhere near your pumpkins happy halloween

Post attachmentPost attachment
0:09
Lena Hall avatar
Lena Hall@lenadroid·Oct 24

some recent patterns I’m seeing talking to founders and business owners - founders open sourcing their products - startups pivoting from b2b saas to hardware, biotech, or media/marketplaces - more extreme polarity of optimism and doomerism

Lena Hall avatar
Lena Hall@lenadroid·Oct 23

When anyone can build anything, how do you make something people actually choose? At AGNTCon I talked about why building something great is only half the job. My favorite part is always the conversations afterward. Grateful I got to share the stage with @techgirl1908 and @sarah_edo. One of the things I talked about is public signal. What people actually understand and remember about your product. We have so much content now, but how much of it actually helps people understand what we’re building? AI summarizes and remixes your content, leaves out important details, or exaggerates small things until customers start hearing promises you never made. And somehow, genuinely different products end up sounding exactly the same. So if your product isn’t converting, maybe you don’t need more content or better distribution. Maybe you need to understand what people are actually hearing about your product, and whether that’s what you intended. This is one of my favorite problems to help teams solve, so let’s chat!

0:47
Lena Hall avatar
⟳ Lena Hall retweeted
Lena Hall@lenadroid·Oct 21

Come with me to AGNTCon! 🤍 Excited to see you there ✨

0:19