Generating structured data from unstructured input is a cornerstone of many AI applications today. Developers leverage the OpenAI API to build sophisticated assistants capable of fetching data, answering complex questions via function calling, extracting structured data for seamless data entry, and enabling multi-step workflows where LLMs can take specific actions.

However, the inherent limitations of LLMs in consistently producing structured output have led developers to employ workarounds such as open-source tooling, intricate prompting techniques, and repeated request retries. These workarounds, while functional, add complexity and compromise efficiency.

OpenAI’s Structured Outputs promises to eliminate these workarounds. It achieves this by constraining OpenAI models to match developer-supplied schemas and by training models to better understand and adhere to complex data structures.

“Structured Outputs solves this problem by constraining OpenAI models to match developer-supplied schemas and by training our models to better understand complicated schemas,” OpenAI said in a blog post.

Internal evaluations using complex JSON schemas have shown remarkable results. The latest model, gpt-4o-2024-08-06, achieved a perfect 100% score in adherence to structured outputs, a significant improvement over the previous gpt-4-0613, which scored less than 40%.

Structured Outputs is offered in two key implementations:

1. Function calling: This method, enabled by setting strict: true within a function definition, allows developers to define the precise structure of data returned by functions called by the model. This feature is compatible with all models supporting tools, including gpt-4-0613, gpt-3.5-turbo-0613, and later versions.
2. Response format parameter: This approach allows developers to provide a JSON Schema via the new json_schema option within the response_format parameter. This is particularly useful when the model needs to respond directly to users in a structured format without invoking tools. Currently, this feature is supported by the latest GPT-4o models: gpt-4o-2024-08-06 and gpt-4o-mini-2024-07-18. Setting strict: true within the response_format ensures the model output conforms to the provided schema.

OpenAI has prioritised safety in Structured Outputs, ensuring it aligns with pre-existing safety policies. The model retains the ability to refuse unsafe requests, indicated to developers by a new refusal string value in API responses. This allows for programmatic detection of refusals, ensuring predictable behaviour and simplified error handling. Notably, the absence of a refusal value and a successful generation process (signified by finish_reason) guarantee a valid JSON output matching the developer-defined schema.

Native support for Structured Outputs has been integrated into OpenAI’s Python and Node SDKs, simplifying its use. Developers can define schemas for tools or response formats by providing a Pydantic or Zod object, which the SDKs automatically convert to JSON Schema. The SDKs also manage the deserialisation of JSON responses into typed data structures and handle potential refusals.

Postbot is Postman’s intelligent assistant designed to tackle two perennial pain points in API development: testing and documentation.

“The major pain points that Postbot solves is allowing you to – by click of a button – generate tests for your APIs, generate documentation,” Chin explained. He added that Postbot can also visualise data, making it easier for developers to interpret large JSON objects.

By tapping into both Postman’s vast knowledge base and a user’s individual workspace, Postbot promises to significantly accelerate development cycles and foster greater collaboration within teams.

AI’s impact on developers: Empowerment, not replacement

Chin firmly dismissed the notion of an “AI apocalypse” leading to mass developer unemployment. Instead, he envisions a future where AI empowers developers to reach new heights of productivity and creativity.

“It’s going to enhance developers’ productivity and it’s going to enhance developers,” Chin emphasised. This collaborative approach, he believes, will be crucial for harnessing the full potential of AI in software development.

While optimistic about AI’s potential, Chin acknowledged concerns surrounding AI code generation, particularly regarding copyright infringement and the use of proprietary data in training models. He stressed Postman’s commitment to ethical AI practices, stating that Postbot is trained exclusively on Postman’s proprietary information and publicly available resources.

Transparency, Chin believes, is paramount in building trust and ensuring responsible AI development. “Open source… we can ensure that we can see the data behind it. That’s why I’ve become a bigger advocate of open-source AI in the last six months.”

The future of API development

Looking ahead, Chin paints an exciting picture of the future API landscape, one where AI acts as a bridge between developers and a world increasingly driven by interconnected services.

“I think we’re going to start to see AIs go out and read someone else’s documentation… As a developer, I could say, ‘Please go connect to this API’,” Chin mused, outlining a future where AI agents streamline API integration, removing technical barriers and unlocking new levels of interoperability.

Chin also teased upcoming developments on Postman’s roadmap, highlighting deeper Postbot integrations and a commitment to becoming a central hub for exploring and evaluating AI APIs.

Practical AI implementation: A look at Postman’s playbook

At this year’s AI & Big Data Expo Europe, Chin will present an “AI/IML playbook” that draws upon Postman’s own experiences in implementing AI solutions. This practical guide will equip attendees with the knowledge and best practices needed to successfully integrate AI into their own workflows and products.

“It’s coming from practicality, the practical implementation of AI here at Postman, going outside of just the toy phase and specifically on how we implemented AI in a way that helps our users and how we went about that,” explains Chin.

With Postbot leading the charge and a clear commitment to responsible AI development, Postman is aiming to empower developers and shape the future of the API landscape for years to come. By addressing key challenges in API testing and documentation, Postman is positioning itself at the forefront of AI-driven development tools.

You can watch our full interview with Sterling Chin below:

Sterling Chin and his Postman colleagues will be at this year’s AI & Big Data Expo Europe to share their expertise. Swing by Postman’s booth at stand #294 to hear more about how solutions like Postbot can improve your workflow.

The latest dispute over App Store rules centres on Apple’s requirement for all iOS app transactions to be processed through their in-app purchase system, a move that Patreon argues undermines their creator-centric philosophy.

In a statement, Patreon explained:

“Patreon exists to let creators make what they love for their biggest fans, and earn an income doing it on their own terms. Everything we build is to make those two things easier for creators. For you.”

Despite Patreon’s commitment to transparency, control, and stability for their creators, they are now facing a deadline of November 2024 to fully integrate Apple’s in-app purchases or risk being removed from the App Store.

This enforced change has significant implications for creators using the platform. Firstly, Apple will be taking a 30% cut of all new memberships and shop purchases made through the Patreon iOS app. Secondly, creators currently utilising first-of-the-month or per-creation billing will be forced to adopt a subscription-based model; the only option supported by Apple’s system.

Patreon emphasises that these changes will not affect existing members, only those who sign up via the iOS app from November onwards. However, for creators accustomed to alternative billing approaches, this shift represents a major disruption to their established workflows.

In an attempt to mitigate the impact of Apple’s 30% fee, Patreon has developed an optional tool that automatically increases prices within the iOS app to offset the cost. While this ensures creators continue to earn the same amount per membership, the decision ultimately rests with them.

Patreon acknowledges the lack of ideal solutions, stating:

“Obviously, neither of these solutions are ideal. But remember, Apple’s fees are only in the iOS app. Your prices on the web and the Android app will remain completely unaffected.”

To ease the transition to subscription billing, Patreon has initiated a 16-month migration process with a deadline of November 2025.  This includes phasing out first-of-the-month and per-creation billing models, a move dictated by Apple’s imposed timeline. 

Patreon is actively supporting creators throughout this process with a range of resources, including personalised assistance for those on per-creation billing, community forums on Discord, and comprehensive FAQs. 

“We didn’t choose this path, but we’re focusing our resources on making subscription billing an even better experience for creators and fans,” Patreon stresses.

Patreon says that it’s committed to continued dialogue with Apple in the hope of finding a more creator-friendly solution. However, even in the courts, previous attempts by other companies have had minimal success.

“In today’s competitive job market, having the right skills is more important than ever, and this data provides clear evidence of the programming languages employers are seeking,” commented Sheldon Chi, ex-Google engineer and creator of System Design School.

“Whether someone is just starting their career or looking to advance, understanding which languages are in demand can significantly improve job prospects and career trajectory.”

The findings highlight the dominance of SQL (Structured Query Language), with 2,291 job postings demanding proficiency in the language. Primarily used for database management, SQL’s advantages extend far beyond the realm of software development. As data analytics becomes increasingly integral to businesses across various sectors, SQL skills are highly transferable and valuable to a wide range of professionals.

Trailing behind SQL in second place is Python, with 1,488 job postings seeking expertise in the language. Python’s popularity stems from its simplicity, versatility, and extensive libraries, making it suitable for a diverse range of applications from web development to data science and machine learning.  

Java secured the third position with 856 job postings. Renowned for its reliability, platform independence (“write once, run anywhere” capability), and extensive use in enterprise applications, Java remains a mainstay in industries spanning finance to Android app development. 

JavaScript, the backbone of front-end web development, took fourth place with 674 job postings. Powering the functionality of countless websites and web applications, JavaScript’s versatility extends to back-end development with frameworks like Node.js, solidifying its status as a crucial skill for full-stack developers.

Top 10 in-demand programming languages:

The research also highlighted the enduring importance of front-end technologies. HTML5 secured fifth place with 572 job postings—its ability to deliver rich multimedia experiences, interactive elements, and responsive designs across devices makes it indispensable for modern web development.

Chi encourages professionals to actively cultivate in-demand skills: “We encourage professionals to focus on these areas to enhance their employability and career growth. By continuously updating skillsets and staying informed about industry trends, people can position themselves as a valuable asset to potential employers.”

The investigations – launched over concerns that both tech giants were unfairly leveraging their dominant market positions, to the detriment of UK app developers and consumers – focused primarily on rules mandating the use of proprietary in-app payment systems. The CMA argued these rules stifle competition, limit consumer choice, and potentially lead to inflated prices. 

Google’s attempts to appease the watchdog with proposed concessions, dubbed ‘Developer-only Billing’ (DOB) and ‘User Choice Billing’ (UCB), proved insufficient. These proposals, intended to offer developers alternative payment options, were deemed inadequate following consultation with industry players. 

App developers – voicing their feedback to the CMA – argued that Google’s concessions fell short, leaving them tethered to the tech giant’s payment infrastructure through hefty commission fees and potentially disruptive ‘pop-up screens’ during transactions.

The CMA’s decision to close the investigations comes amidst the recent enactment of the DMCCA. This new legislation equips the watchdog with enhanced powers to address competition concerns within the digital marketplace. Crucially, it paves the way for the designation of companies like Google and Apple as having ‘strategic market status’ in areas like mobile ecosystems. 

“Once the new pro-competition digital markets regime comes into force, we’ll be able to consider applying those new powers to concerns we have already identified through our existing work,” said Will Hayter, Executive Director for Digital Markets at the CMA.

This shift in approach suggests the CMA is poised to deploy its expanded arsenal should either tech titan be designated with ‘strategic market status’. This designation would grant the CMA a broader purview to address competition issues holistically, potentially leading to more comprehensive interventions.

“It’s critical that tech businesses in the UK, including app developers, can have access to a fair and competitive app ecosystem, helping to grow the sector, boost investment, and result in better outcomes for UK consumers,” added Hayter. “These are all factors we are considering before launching our first investigations under the new regime.”

The CMA’s decision signals a new chapter in the ongoing tussle to ensure fairness and competition within the app economy.  With its sights set on the DMCCA’s potent toolkit, the watchdog aims to tackle any perceived market imbalances head-on.

Alongside security patches, the update delivers bug fixes, minor feature enhancements, and changes to the platform.

The most pressing issue tackled by this update is a critical vulnerability (CVE-2024-6800) impacting instances employing SAML SSO with specific Identity Providers (IdPs).

CVE-2024-6800 was discovered through GitHub’s Bug Bounty programme and could allow an attacker to forge a SAML response, potentially granting them access to user accounts with site administrator privileges.

This release also addresses two medium-severity vulnerabilities:

• CVE-2024-7711: This vulnerability allowed attackers to modify the title, assignees, and labels of issues within public repositories. Private and internal repositories remained unaffected.
• CVE-2024-6337: Attackers could exploit this vulnerability to expose issue content from private repositories using a GitHub App with specific read and write permissions. It’s important to note that this exploit required a user access token and did not impact installation access tokens.

Beyond security fixes, 3.13.3 brings several notable changes:

• Enhanced visibility: Users gain increased visibility into the state of gists, networks, and wikis with the addition of app state information within the spokesctl info output. Additionally, the spokesctl check command can now diagnose and often rectify empty repository networks.
• Improved stability and performance: Several bug fixes target issues related to hotpatching, configuration updates, and database migrations, resulting in improved system stability.
• Usability improvements: Administrators benefit from more granular control over the maximum object size within repositories. Users can now customise their link underline styling preferences within the accessibility settings.

While this update enhances security and stability, GitHub acknowledges several known issues outlined within the official release notes. These include potential errors during configuration runs, issues with audit log data migration, and increased memory utilisation.

To review the full list of changes, please refer to the official release notes on GitHub’s website.