OpenAI released a preview of new 'reasoning' models

FIRST PASS

OpenAI introduces a new series of ‘reasoning models for solving hard problems’ (OpenAI)

OpenAI just released a preview of a new series of AI models designed to excel at complex reasoning tasks in science, coding, and math.

Key points:

• Like human reasoning, the o1-preview model spends more time thinking through problems before responding.

• In tests, it significantly outperformed previous models, scoring 83% on International Math Olympiad qualifying exams (compared to GPT-4o's 13%).

• A smaller, faster version called o1-mini is also being released, optimized for coding tasks.

• Access will be rolled out gradually, starting with ChatGPT Plus and Team users, and API tier 5 developers.

• OpenAI has implemented new safety measures and formalized agreements with U.S. and U.K. AI Safety Institutes.

Merlin’s Notes: Looks like the ‘strawberry’ model rumors from a few weeks ago were true. It can certainly be a bit frightening to see the pace at which AI is advancing, but it’s no excuse to ignore it.

While it’s too soon to tell what kind of impact these models will have on use cases and broader industries, we’d like to speculate here:

1. Leveling the Playing Field: With AI coding assistants emerging as the ‘killer’ app (as we highlighted previously), we might see a further leveling of the playing field in software development. This could lower barriers to entry for tech startups, potentially increasing competition for established players.

2. R&D Acceleration: Enhanced reasoning capabilities could significantly speed up R&D processes, particularly in STEM fields, which might lead to faster innovation cycles and potentially disrupt traditional R&D-heavy industries.

3. True Automation of High-Skill Tasks: As these models become more capable, it seems inevitable that we’ll see more automation of tasks handled by even highly skilled professionals. Think AI analysts reviewing CIMs, conducting market research, and preliminary diligence as well or better than a human.

For PE, all of this underscores the need to reassess the AI readiness of portcos, identify new use cases, and stay on top of trends that could impact fund operations & potential investment opportunities.

Klarna ditches Salesforce and Workday, betting big on AI for operational efficiency (SeekingAlpha)

Swedish fintech giant Klarna is cutting ties with major software providers as it pivots towards AI-driven solutions, aiming for significant cost savings.

Key points:

• Klarna has shut down Salesforce for CRM services and plans to shut down Workday within weeks.

• Klarna is developing in-house AI solutions to replace these third-party services.

• The changes are expected to save tens of millions of dollars annually.

• CEO Sebastian Siemiatkowski says AI is enabling a more lightweight tech stack with higher-quality operations.

Merlin’s Notes: While we’re skeptical of how this will play out, the move signals a potentially seismic shift in how companies think about traditional enterprise software solutions in the AI age.

The decision to replace established SaaS providers with in-house AI solutions could have far-reaching implications for the tech industry and PE investments.

It's important to note that companies have tried this in the past and it's safe to say that building a full-scale Enterprise SaaS platform is easier said than done. But with recent advances in AI coding capabilities, building lightweight internal tooling is certainly more feasible.

Is Klarna starting a trend that will slowly destroy the traditional SaaS model or will this result in a massive miscalculation? Only time will tell.

In any case, we think this is worth paying attention to because:

1. There’s likely potential for substantial cost savings for internal operations and portco operations through AI adoption

2. It underscores the importance of either in-house or external tech capabilities to take full advantage

3. There could be possible disruption in the enterprise software market as companies scrutinize their tech spend
   

JARGON BUSTER

USE CASE SPOTLIGHT

Goldman intensifies focus on project Legend thanks to AI

Goldman Sachs developed Legend, a centralized hub for all its important data about a decade ago, and is now reprioritizing it for the GenAI age as its data continues to explode.

Key features of Legend:

• Centralizes all of Goldman's important data in one platform

• Provides a single access point for employees across different teams

• Enables quick discovery of connections that could lead to multimillion-dollar deals

• Facilitates automation of operations-heavy work

• Supports AI model building for pattern recognition

Impact and use cases:

1. Business growth: Legend is being leveraged to grow Goldman's asset- and wealth-management business, as well as its sales and trading operations

2. Operational efficiency: By providing "one version of the truth," Legend reduces the need for constant data reconciliation and processing across systems

3. Cost savings: Consolidation of data infrastructure leads to reduced operational costs

4. AI and automation enhancement: The platform supports the development of generative AI models and automates back-end processes

The project, part of Goldman's broader AI strategy, aims to free up bankers' time for higher-value tasks such as client interactions and strategic thinking.

TOOL OF THE WEEK

Replit

Replit is an online integrated development environment (IDE) that allows users to write, run, and collaborate on code in various programming languages directly in a web browser.

The platform has gained tremendous popularity among developers, students, and educators for its ease of use and robust features that make writing and learning code easier than ever before.

Key features:

• In-Browser coding: Write and execute code without the need for local installations

• Replit AI: Use AI to turn natural language into entire apps

• Collaboration Tools: Real-time code sharing and pair programming capabilities

• Hosting: Deploy web applications directly from the Replit environment

• Educational Resources: Built-in tutorials and coding challenges for learners

• Cursor Integration: Seamless connection with cursor for enhanced development workflow, allowing developers to leverage the power of the Cursor IDE, while hosting live and seeing instant changes to the front-end

• Instantaneous Syncing: Real-time code updates between Replit and Cursor, or VS code, enabling rapid development and testing across platforms.