
Head of Developer Growth at @ActianCorp. Prev @Massive_Com @Box. I make developer content with a focus on AI. Views are my own.
Austin, TX
Joined January 2021
- Tweets256
- Following157
- Followers130
- Likes283
Agreed. Should not be a hard or rare thing to solve. What’s the hold up? Does not encourage me to want to use driverless cars going forward. A human would’ve answered their phone or turned around and given them back. 1/10 experience. 👎
Replying to @robotaxi
1
31
Alex Novotny retweeted
DocuSign wants $8,100/yr from a 15-person team just to sign PDFs. So I built SignuDoc: open source, one-click deploy to your own @vercel account, $0 per seat.
White-label branding, full audit trails, tamper-evident seals and your documents never leave your cloud.
Alex Novotny retweeted
Stop trusting your hit rate.
First, a hit only means two questions scored close enough. Our semantic cache hit 79.2%, and 39 of those hits were wrong answers.
Second, a changed word or number can score as high as a paraphrase. “Should I not cancel my card?” matched “Should I cancel my card?” at 0.990.
Third and fourth, another tenant’s answer and an outdated answer both count as hits.
We ran seven tests, and filters and targeted checks cut the wrong answers the threshold let through. Before you ship your cache, measure its false-positive rate next to its hit rate 👇 bki.sh/MzOUing
Alex Novotny retweeted
You asked for it, we delivered.
Field selection and Apache Arrow responses are live on supported endpoints.
Ask for the specific fields you use. Get them back as a columnar Arrow table instead of a JSON blob.
learn more below:
Alex Novotny retweeted
Introducing Jaspersoft Desktop, now available for early access! Describe the report you want, Jaspersoft Desktop builds it, and your report data stays local. Start your 60-day trial with a free Jaspersoft Community Account: bit.ly/4AcgM3h
#EmbeddedAnalytics
Alex Novotny retweeted
We grew a Pi 5 agent's memory 100x, to 100,000 vectors. Median recall moved from 1.90 ms to 4.17 ms. A 2.2x cost for 100x the data, all on the board 🫡
Alex Novotny retweeted
Is this the best way to build self-evolving agents with GPT-6 Astra?
Astra has a 1.05 million-token context window that resets with every new session.
We gave it an external memory. It stores every verified fix, and Astra checks that memory before touching new code.
Watch this before building an agent workflow with GPT-6 Astra! bki.sh/QwKTRsI
Alex Novotny retweeted
5 things that happened in AI infrastructure last week that'll affect how you build.
1️⃣ The agent loop is becoming its own infrastructure layer. What you build on top of it matters more than which model sits inside it.
2️⃣ Vector search alone can't handle production agent workloads. An agent that gets an exact lookup wrong loses the user. Hybrid retrieval fixes that.
3️⃣ Several vendors shipped disk-based indexes this week. If cost and memory weren't in your infrastructure decision before, they should be now.
4️⃣ Agents that place calls and touch codebases are surfacing a question teams haven't answered yet: who owns the data trail those agents leave behind?
5️⃣ Air-gapped facilities, regulated networks, and disconnected environments can't run cloud-hosted agents. If that's your deployment environment, your infrastructure shortlist just got shorter.
What are you watching out for this week?
I'm 35.
Developer Advocate from a place called Oklahoma. Currently based in Austin.
Looking to connect with more software developers, developer advocates & people building with AI. 🤠
+1 if you play boards too. ♟️
I can barely get my Tesla to miss a pothole on the road, and IMHO thats the closest thing to real world AGI we have right now. I think we are fine. Everyone needs to chill out.
Jacob is correct here—we really do earnestly believe AI could kill all humans! I personally think it is >10% within the next decade. I believe Anthropic is trying its best, but we do not yet have a plan to solve alignment for superintelligence and are not clearly on track to.
1
86
Alex Novotny retweeted
Thanks @andre_z.
We respect what the @qdrant_engine team has built. Their API is clean and deliberate. We made our API compatible because it’s a standard developers already know. Low friction, easy migration, better for the whole space.
But under the hood, VectorAI DB is fundamentally different. We’re running on Actian edge database technology, a battle-tested database engine with decades of maturity, paired with our own indexing orchestration. The performance gains we quoted are from standard benchmarks like VectorDBBench.
We’re here for the engineering conversation.
Replying to @ActianCorp
@ActianCorp launched a vector DB. Their comparison post says it's 22x faster than Qdrant.
I was curious what they'd built, so I opened the docs...
Core concepts: Collections, points, vectors, payload... Wait. I know it from somewhere. 🤔
Alex Novotny retweeted
A write up of some recent work we did @massive_com to tag 8-K filings using an LLM pipeline. Probably the most interesting learning was that asking LLMs for a quality score of their own tag was terribly calibrated, but using a fresh context call did much better!
Why does he look like cottage cheese as a person? 😂
Introducing Aidan, your newest AI employee
Aidan is the first computer-using AI built for realtime conversation
We’ve raised $45M from @sequoia and @8VC to bring Aidan to the world
@NotionHQ and @DecagonAI already use Aidan to run customer interactions. This is how @sablehq works:
2
4
265
Alex Novotny retweeted
Earnings aren't coin flips.
@earnings_watch built a platform to prove it, on Massive data and 1,000+ options traders use it daily.
Here are the Massive products they use:
→ Options Contracts + snapshots power the earnings simulator and VolScanner
→ Fair Market Value cleans up the rankings
→ Stocks Aggregates drive post-earnings drift analysis
One API pattern across every tool. No mismatched data between screening, simulating, and analyzing.
Read more in the case study.
massive.com/blog/case-study-…
Anyone else try making a pet 🦉 in ChatGPT? Here's mine. I think he's pretty cute? What do you think? I'm such a sucker for adorable little guys like this, even if I used up all my Codex usage to make him. 🫠
GPT 5.6 Sol ☀️ did really well here too. 👏
2
152
😲🤯Another day another model release.
Just tested GPT-5.6 Sol on the new @massive_com Remote MCP Server with this monster query…
Using the massive (pun intended) Massive data, it analyzed the latest FOMC rate decision’s impact on QQQ + options:
- 1-min QQQ bars from 30 min before the statement to 90 min after the press conference
- One liquid 25-45 DTE expiration
- ATM call/put + ~5% OTM call/put
- Price & volume reaction, bid-ask spreads, IV changes, downside skew, gamma exposure
- Statement vs press conference: which one actually moved the market more?
- Plus the exact graphs requested
No code. No manual data pulls. Just one prompt → full professional-grade event study in seconds complete with a graph. 📈📊
Woah 🤯 I connected @spacexai's Grok 4.5 to @massive_com using their new remote MCP server - asked it to analyze @Tesla performance and create a powerpoint with the results. Did it all in under two minutes. Pretty blown away by this. This is just the tip of iceberg too. I haven't even tried my hard queries yet.






















