 |
|
PULSE NEWSLETTER
|
|
Trustworthy AI, temporal integrity, AI-ready market data, and how good LLMs really are at writing q.
|
|
|
|
| |
|
Hi ala,
In this issue of KX Pulse we explore a question that's getting harder to ignore. How much can capital markets trust their AI outputs?
It starts with the case for giving AI decisions a temporal flight recorder that can reconstruct what a system knew at the moment it acted.
Ashok Reddy then joins the podcast on why ground truth lets you test an output against what the market actually did, followed by Mick Hittesdorf on what it really takes to make market data AI-ready.
From there we look at how far coding agents have come at writing q, and a new way to pull historical market data straight into KDB-X.
Grab a coffee and enjoy! ☕
|
|
|
|
|
|
|
Why Capital Markets AI Needs a Temporal Flight Recorder
|
|
|
|
In his latest blog, our CEO Ashok Reddy explains how Temporal AI Infrastructure can preserve a time-accurate record of the data, models, rules, permissions, actions, and outcomes surrounding a decision, giving teams what they need to reconstruct the system state that existed at a specific point in time.
| ▪ |
Preserve event time and knowledge time: Bitemporal data records when something happened and when the system knew about it, helping teams avoid using information that was unavailable at decision time. |
| ▪ |
Record the full decision state: Data, model versions, controls, permissions, actions, and outcomes need a common temporal history. |
| ▪ |
Support reconstruction: Risk, compliance, model validation, and engineering teams can review the information and controls available when a decision was made. |
|
| Read the full blog → |
|
|
|
|
Ashok Reddy on Temporal Integrity and Trustworthy AI in Capital Markets
|
|
|
|
AI is probabilistic, which makes ground truth critical when firms need to decide whether an output deserves confidence. In this episode of the KX Pulse podcast, Ashok Reddy looks at how firms can test AI outputs against observable market evidence, and why preserving temporal context matters when you need to reconstruct what the system knew at the point a decision was made.
| ▪ |
Identify lookahead bias: Recording when information became available helps teams detect historical tests that use data unavailable at the time. |
| ▪ |
Reproduce decision context: Firms can test whether they can recreate the data and system state behind a specific AI output. |
| ▪ |
Test signals against outcomes: Comparing AI-generated signals with subsequent market behaviour gives firms a clearer way to assess whether those signals held up. |
|
| Watch the podcast → |
|
|
|
|
Mick Hittesdorf on AI Data Readiness in Capital Markets
|
|
|
|
AI workflows depend on the quality and structure of the market data supplied to them. Mick Hittesdorf, Product Architect for OneTick products, joins the KX Pulse podcast to look at what makes market data genuinely ready for AI, and how structure, granularity, metadata, temporal context, and validation shape how much confidence firms can place in the result.
| ▪ |
Structure data for the workload: Models and agents need market data at the right level of granularity for the task. |
| ▪ |
Keep the context attached: Timestamps, identifiers, corporate actions, and metadata shape how market events are interpreted. |
| ▪ |
Prove the data is fit for use: Repeatable validation helps teams assess accuracy, consistency, and timeliness before data enters an AI workflow. |
|
| Watch the podcast → |
|
|
|
|
How Good Are LLMs at Writing q?
|
|
|
|
In his latest blog, Scott Rich looks at how far coding agents have come in writing q, using results from KX's Q Evaluation Harness (QEval). Frontier models paired with coding agents solve more than 85% of QEval problems, while the leading configuration exceeds 95%, and he unpacks where q-specific guidance helps and where experienced developers still need to make the calls.
| ▪ |
Execution improves code generation: Agents that can run and debug q code perform better than one-shot generation. |
| ▪ |
q-specific guidance matters: Language-specific instructions help agents avoid common errors and follow established q development practices. |
| ▪ |
Expertise still matters: Developers remain responsible for architecture, system design, and complex technical judgment. |
|
| Read the full blog → |
|
|
|
|
Access Historical Market Data From KDB-X
|
|
|
|
Historical market-data research can become an infrastructure project before the analysis even starts. In this on-demand webinar, Peter Simpson shows how to query OneTick Market Data directly from a q session, run OneTick SQL against the historical data, and bring the results back into KDB-X as q tables.
| ▪ |
Query history on demand: Retrieve specific markets, symbols, date ranges, and analytical views for analysis in KDB-X. |
| ▪ |
Work with market microstructure data: The session covers trades, quotes, NBBO data, order book depth, and consolidated market views. |
| ▪ |
Account for historical context: The workflow covers corporate actions, futures contracts, symbology, and other time-dependent factors that can affect analysis. |
|
| Watch the webinar → |
|
|
|
|
KX Strengthens Its Capital Markets Leadership Team
|
|
|
|
A warm welcome to James Hollands, who has joined KX as Chief Revenue Officer, bringing more than 20 years of capital-markets technology experience. James joins Matthew Bagley, Chief Financial Officer, and Adam Bozek, General Counsel, who were also appointed earlier this year.
|
| Read the announcement → |
| |
|
|
Where to Find Us
|
|
Events & Workshops
OneTick Workshop — Tuesday 15 September
Join Peter Simpson for a practical workshop on market data and trading analytics, working with research-ready data and OneTick analytics workflows.
TradeTechFX, Amsterdam — Wednesday 16 September
Alex Weinrich joins the Trading Intelligence Interview: turning fragmented data into better routing, timing, and liquidity decisions for the modern FX desk.
Data Management Summit, New York City — Thursday 17 September
Mick Hittesdorf joins the panel "Scaling with trust — data quality engineering as a driver of business performance", covering AI data quality, RAG, anomaly detection, observability, and data drift.
FISD New York — Thursday 17 September
KX is sponsoring FISD New York at 7 World Trade Center, bringing together market data consumers, vendors, exchanges, and other industry participants.
Tech & Data in Financial Markets — Thursday 24 September
KX speaks in New York as senior leaders discuss AI, data strategy, governance, technology modernization, risk, and market structure.
|
| |
|
| |
|
Join Our Community
Applications are still open for our Developer Advocacy Program, 'Community KXperts'! This program is ideal for anyone passionate about sharing their knowledge on KX through blogs, articles, or other content. To apply, contact evangelism@kx.com.
|
| |
|
|
|
| KDB-X | Why KX? | Resources | Developers |
|
|
This email was sent to ala@niepodam.pl by: KX, 45 Broadway, New York, NY 10006, USA
Manage Preferences or Unsubscribe | Privacy Notice |
|
