The most important AI developments from around the world, summarized by AI so you stay informed in minutes.
Last updated: 18/4/2026, 6:45:05 am (IST)
🧠
AI Shifts & Debates
AI News Daily Top 5
2026-04-18
AIDays.in
- OpenAI faces another leadership exodus as executives depart.
- Advances in AI are highlighted by new robot brains and improved RAG.
- The AI market is being questioned, with some seeing inflated demand.
- Tech companies are vying for government AI contracts.
01
OpenAI loses multiple executives in latest leadership shakeup
Such instability at a leading AI firm like OpenAI could impact its pace of innovation and strategic direction, with potential ripple effects across the global tech landscape.
CNBC Tech
02
Stop using naive RAG – adding relationships to AI context
This advancement moves RAG beyond simple keyword matching towards a more intelligent understanding of complex information landscapes, improving AI's effectiveness in technical domains.
Hacker News (AI)
03
AI chipmaker Cerebras files to go public after scrapping IPO plans last year
Cerebras's IPO signals a potential major liquidity event for a key enabler of advanced AI development, highlighting the ongoing consolidation and investment in the foundational AI hardware sector.
CNBC Tech
04
Trump says he had 'no idea' Anthropic's Amodei met with White House about Mythos
This situation highlights the complex and often opaque relationship between major AI developers and political entities, even amidst past governmental sanctions.
CNBC Tech
05
Kevin Weil and Bill Peebles exit OpenAI as company continues to shed ‘side quests’
This pivot suggests OpenAI is prioritizing monetization and practical business applications for its AI advancements, potentially impacting the future direction and accessibility of its consumer technologies.
OpenAI is facing another leadership exodus, with multiple executives departing the AI powerhouse. This comes hot on the heels of their product and business chief, Fidji Simo, beginning a medical leave. The repeated departures signal ongoing internal turbulence at the company, which remains a dominant force in AI development.
Key Takeaways
OpenAI is experiencing a significant leadership churn.
The departures follow closely behind the medical leave of their product and business lead.
This shakeup raises questions about stability and future direction at OpenAI.
Why it matters: Such instability at a leading AI firm like OpenAI could impact its pace of innovation and strategic direction, with potential ripple effects across the global tech landscape.
A new tool called Mindex is addressing a key limitation in Retrieval Augmented Generation (RAG) systems: their inability to preserve relationships between retrieved information. Unlike traditional RAG which fetches isolated text chunks, Mindex integrates semantic search with a knowledge graph. This allows it to not only find relevant documents but also explicitly surface how these documents and entities within them connect, crucial for complex domains like software documentation and infrastructure. It's designed for seamless integration into developer workflows via CLI and MCP, compatible with tools like Claude Code and Cursor.
Key Takeaways
Naive RAG struggles to maintain contextual relationships between retrieved information.
Mindex introduces a knowledge graph layer to RAG for enhanced relational understanding.
The tool is developer-centric, offering CLI/MCP integration for popular AI coding tools and agents.
Why it matters: This advancement moves RAG beyond simple keyword matching towards a more intelligent understanding of complex information landscapes, improving AI's effectiveness in technical domains.
AI chip giant Cerebras has refiled for an IPO, a year after shelving its initial public offering plans. The company, known for its wafer-scale AI processors, revealed a strategic partnership with OpenAI, including a warrant for OpenAI to purchase Cerebras stock. This move signals Cerebras's confidence in its growth trajectory and its deep integration with a leading AI development firm.
Key Takeaways
Cerebras is once again pursuing an IPO, indicating renewed market optimism and business progress.
A significant partnership with OpenAI, including equity warrants, underscores Cerebras's strategic importance to leading AI players.
The company's wafer-scale AI chips are critical infrastructure for demanding AI workloads, as demonstrated by the OpenAI collaboration.
Why it matters: Cerebras's IPO signals a potential major liquidity event for a key enabler of advanced AI development, highlighting the ongoing consolidation and investment in the foundational AI hardware sector.
Former President Donald Trump has publicly stated he was unaware of a meeting between Anthropic CEO Jack Clark (note: the article states Amodei, but Clark is the commonly known CEO of Anthropic, and the context implies a high-level executive meeting) and White House officials regarding their AI model 'Mythos.' This comes as a surprise, given that Trump had previously blacklisted Anthropic's AI, Claude, just two months prior. The meeting suggests a potential shift or at least a re-engagement with Anthropic by political circles, despite prior governmental actions.
Key Takeaways
Donald Trump claims ignorance regarding Anthropic's executive meeting with the White House.
The meeting concerning Anthropic's AI 'Mythos' occurred despite a prior 'blacklist' of their Claude model by Trump.
This development hints at potential dialogue or reconsideration of AI vendors by political entities.
Why it matters: This situation highlights the complex and often opaque relationship between major AI developers and political entities, even amidst past governmental sanctions.
OpenAI is undergoing a significant strategic shift, with high-profile departures including Kevin Weil and Bill Peebles coinciding with the shutdown of its Sora video generation project and the integration of its science team. This move signals a decisive pivot from ambitious consumer-facing 'moonshots' towards a laser focus on enterprise AI solutions. The company is evidently streamlining its operations by cutting 'side quests' to concentrate on core business applications.
Key Takeaways
Key OpenAI personnel Kevin Weil and Bill Peebles have departed.
OpenAI is discontinuing its Sora video generation project.
The company is integrating its science team and refocusing on enterprise AI.
This represents a strategic shift away from consumer-focused projects towards B2B applications.
Why it matters: This pivot suggests OpenAI is prioritizing monetization and practical business applications for its AI advancements, potentially impacting the future direction and accessibility of its consumer technologies.
OpenAI is seeing another high-profile departure with Kevin Weil, formerly of Instagram and a key leader at the ChatGPT-maker, announcing his exit. Weil's AI science application division will be integrated into the broader Codex project. This move follows a period of significant internal restructuring and talent shifts within OpenAI.
Key Takeaways
OpenAI's Kevin Weil, a former Instagram VP, is leaving the company.
Weil's AI science application unit will be merged into the Codex team.
This departure is part of ongoing organizational changes at OpenAI.
Why it matters: This leadership change could signal shifts in OpenAI's strategic focus and internal operations, impacting its AI development trajectory.
A CNBC Tech analysis suggests that the current 'explosive' demand for AI, often measured by 'tokens,' might be significantly inflated. The article singles out Anthropic as the only major player appearing realistic about AI's actual usage metrics. This challenges the prevailing narrative of unrestrained AI growth, implying a potential overestimation of its current market penetration and impact.
Key Takeaways
AI demand metrics, particularly 'tokens,' may be artificially inflated.
Anthropic is highlighted as an outlier for presenting a more grounded view on AI usage.
The article questions the current scale of AI adoption and its explosive growth projections.
Why it matters: This perspective could influence investment decisions and strategic planning within India's burgeoning tech sector, potentially tempering expectations around AI-driven ventures.
Google DeepMind's new Gemini Robotics-ER 1.6 is enhancing robot capabilities, particularly in planning and precise action execution. A key advancement is its improved ability to interpret readings from measuring instruments, leading to more accurate real-world interactions. This upgrade signifies a move towards more intelligent and context-aware robotic systems.
Key Takeaways
Gemini Robotics-ER 1.6 boosts robot planning and action precision.
New capability allows robots to accurately read measuring instruments.
Enhances robots' perception and interaction with physical tools.
Why it matters: This development paves the way for robots to perform more complex tasks in environments requiring fine-grained measurement and control, crucial for manufacturing and logistics in India.
Google is further integrating AI into its Chrome browser, a move that could significantly alter how users interact with web content. Soon, websites might appear directly alongside AI-generated responses, effectively bypassing the need for a traditional page visit. This development, highlighted by The Decoder, signals Google's intent to make direct website navigation less essential for information retrieval.
Key Takeaways
Chrome's AI integration is advancing to display websites alongside AI answers.
This feature aims to reduce the necessity of direct website visits.
Google is prioritizing AI-driven information delivery within the browser experience.
Why it matters: This shift has significant implications for website traffic, ad revenue models, and the fundamental way users discover and consume online information, particularly impacting content creators and businesses reliant on direct engagement.
The US Federal Aviation Administration (FAA) is piloting a new artificial intelligence tool designed to enhance air traffic management, with tech giants Palantir Technologies and Thales SA, alongside Air Space Intelligence Inc., locked in a competitive bid. This initiative signals a significant push by the FAA to leverage AI for improving the safety, efficiency, and capacity of the National Airspace System. The competition among these prominent players highlights the growing importance of AI solutions in critical infrastructure sectors.
Key Takeaways
FAA is actively exploring AI for air traffic control.
Palantir and Thales are key contenders in this competitive bid.
The initiative aims to modernize air traffic management systems.
Why it matters: This development underscores the increasing role of AI in ensuring the robustness and efficiency of critical national infrastructure, with potential implications for global aviation standards.
Anthropic CEO Dario Amodei is reportedly set to meet with White House Chief of Staff Susie Wiles to resolve a protracted dispute with the Pentagon. The crux of the negotiation appears to be Anthropic's new Mythos AI model, which boasts advanced cybersecurity capabilities. This meeting suggests the US federal government is seriously considering leveraging Mythos, potentially overcoming previous reservations.
Key Takeaways
Anthropic's Mythos AI model is under consideration for federal government use.
Mythos's advanced cybersecurity features are a key factor in the potential adoption.
A high-level meeting at the White House indicates a softening of previous Pentagon reservations.
Why it matters: This situation highlights the growing recognition of advanced AI's strategic importance for national security and the complex negotiations involved in its government deployment.
Alibaba's latest open-source LLM, Qwen3.6-35B-A3B, is making waves by outperforming Google's Gemma 4-31B on agentic coding and reasoning benchmarks. Despite activating only a fraction of its 35 billion parameters at any given time, Qwen3.6 demonstrates superior performance, suggesting a novel and efficient approach to large model architecture.
Key Takeaways
Alibaba's Qwen3.6-35B-A3B surpasses Google's Gemma 4-31B in coding and reasoning tasks.
Qwen3.6 utilizes a sparse activation mechanism, engaging only 3 out of 35 billion parameters simultaneously.
This efficiency leads to stronger performance on specific AI benchmarks, challenging larger, denser models.
Why it matters: This development highlights advancements in efficient LLM design, potentially paving the way for more performant and accessible AI models for developers and researchers.
#AI#Large Language Models#Open Source#Alibaba#Google Gemma#Benchmarks#India Tech
GitHub's recent blog post details the process of building an emoji list generator using the GitHub Copilot CLI. The demonstration took place during a 'Rubber Duck Thursday' stream, showcasing practical application of AI-assisted coding for a specific development task. This highlights how tools like Copilot CLI can streamline the creation of custom utilities and developer workflows.
Key Takeaways
Demonstrates practical use of GitHub Copilot CLI for utility development.
Highlights 'Rubber Duck Thursday' as a platform for live AI coding demonstrations.
Explains how Copilot CLI can assist in generating lists and custom tools.
Why it matters: This showcases how AI coding assistants are evolving beyond simple code completion to assist in building functional developer tools, potentially boosting productivity for engineers in India and globally.
US officials from the Trump administration met with Anthropic CEO Dario Amodei to discuss granting them greater access to the company's advanced AI model, 'Mythos.' The White House described the meeting as 'productive and constructive.' This engagement signals a growing interest from government entities in leveraging cutting-edge AI capabilities for their purposes.
Key Takeaways
Trump administration officials are actively seeking access to Anthropic's 'Mythos' AI model.
The meeting between the administration and Anthropic CEO was characterized as positive.
Government interest in advanced AI models like Mythos is on the rise.
Why it matters: This interaction highlights the increasing strategic importance of leading AI models and potential government efforts to control or utilize them, a trend with significant implications for the AI landscape in India and globally.
The AI chatbot landscape is rapidly shifting, with OpenAI's ChatGPT experiencing a significant market share decline. Anthropic's Claude has seen phenomenal growth, doubling its share in just one month and surpassing competitors like Deepseek and Grok. While ChatGPT remains dominant, Google Gemini is emerging as a major challenger, now accounting for a quarter of all AI traffic.
Key Takeaways
ChatGPT is losing market share at an accelerated pace.
Claude is experiencing explosive month-over-month growth, rapidly gaining ground.
Google Gemini is becoming a significant player, capturing a substantial portion of AI traffic.
Why it matters: This indicates a highly competitive and dynamic AI market where established leaders are facing increasing pressure from agile challengers, potentially leading to more diverse and advanced AI offerings.
This Towards Data Science article explores moving beyond basic prompt engineering by leveraging 'agent skills' within AI workflows for data science. The author details how they transformed an eight-year-old weekly data visualization routine into a reusable, automated AI process. This approach emphasizes building more sophisticated and adaptable AI systems that can perform complex data science tasks autonomously.
Key Takeaways
AI 'agent skills' offer a more advanced paradigm than simple prompting for data science tasks.
Complex data analysis and visualization routines can be automated and made reusable using AI agents.
This shift enables more efficient and scalable data science operations by reducing manual intervention.
Why it matters: Embracing agent skills in AI is crucial for unlocking the next level of automation and intelligence in data science, particularly for recurring analytical tasks.
Hugging Face's latest blog post details their approach to building a high-performance, multilingual Optical Character Recognition (OCR) model. The key innovation lies in the extensive use of synthetic data generation to overcome limitations in real-world datasets, enabling the model to accurately process text across numerous languages. This method significantly accelerates training and improves robustness, especially for low-resource languages.
Key Takeaways
Synthetic data generation is a powerful technique to overcome data scarcity in OCR model training.
The developed multilingual OCR model demonstrates improved speed and accuracy across diverse languages.
This approach is particularly beneficial for languages with limited readily available OCR training data.
Why it matters: This research contributes to making advanced OCR technology more accessible and effective for a wider range of languages, potentially democratizing information access and digital transformation in India.
The AI landscape is experiencing a widening 'AI Anxiety Gap' between industry insiders and the general public, marked by significant spending, growing suspicion, and the emergence of new jargon. OpenAI's aggressive acquisition strategy, spanning finance apps to talk shows, alongside companies like a rebranded shoe company focusing on AI infrastructure and Anthropic's announcement of an 'too powerful to release' model, highlight this escalating divergence. This rapid development and strategic maneuvering are creating a perception of exclusivity and potential unease among those outside the core AI development sphere.
Key Takeaways
A growing 'AI Anxiety Gap' exists between AI insiders and the public, characterized by increased spending, suspicion, and new terminology.
OpenAI is actively acquiring various companies, indicating a broad strategic expansion beyond core AI research.
Concerns over AI safety and power are leading to the deliberate withholding of certain advanced models, such as by Anthropic.
Why it matters: This widening gap and the aggressive, opaque moves by major AI players risk alienating the public and fueling anxieties about AI's societal impact and control.
This Wired AI piece warns that AI's infiltration into newsrooms, marketed as a boost to efficiency, presents a significant trade-off that publishers may be downplaying. The core concern is the potential erosion of human authorship and journalistic integrity as AI tools increasingly take on the drafting of stories. The author strongly resists AI's role in their own creative process, highlighting a fundamental clash between automated content generation and the essence of original storytelling.
Key Takeaways
AI is being adopted in newsrooms for perceived efficiency gains in content creation.
There's a growing concern about the profound implications of AI-assisted writing on journalism.
The article expresses strong opposition to AI drafting personal stories, emphasizing human authorship.
Why it matters: This development raises crucial questions about the future of original content, the role of human creativity in journalism, and the potential for AI to redefine authenticity in storytelling.
A new approach discussed on Towards Data Science suggests that robust classification models can be achieved with surprisingly few labeled data points. This research explores how unsupervised learning techniques can be leveraged to significantly reduce the reliance on extensive manual labeling, a common bottleneck in AI development. The core idea is to utilize the power of unsupervised learning to extract meaningful representations from unlabeled data, which can then be effectively fine-tuned with a minimal set of labels.
Key Takeaways
Significant reduction in labeling effort for AI model training.
Leveraging unsupervised learning to boost classifier performance.
Potential for faster and more cost-effective AI deployment.
Why it matters: This research has the potential to democratize AI development by lowering the barrier to entry for creating high-performing classifiers, especially in data-scarce scenarios prevalent in many Indian industries.
Meta's new Just-in-Time (JiT) testing approach is revolutionizing bug detection in AI-assisted software development. By dynamically generating tests during code reviews, leveraging LLMs and mutation testing, Meta has seen a significant ~4x improvement in identifying bugs compared to traditional static test suites. This shift towards change-aware, AI-driven testing aligns with the evolving landscape of agentic development environments.
Key Takeaways
Meta's Just-in-Time (JiT) testing dynamically generates tests during code review.
This AI-driven method, using LLMs and mutation testing, improves bug detection by approximately 4x.
The approach is crucial for agentic development environments that prioritize change-aware testing.
Why it matters: This represents a significant leap forward in making AI-powered software development more robust and efficient by catching bugs earlier in the development cycle.
KDnuggets features an open-source guide for building an AI-powered customer sentiment and topic analyzer for call recordings. The tool leverages OpenAI's Whisper for speech-to-text transcription, BERTopic for identifying key themes, and Streamlit for an interactive dashboard. This allows businesses to gain actionable insights from customer interactions.
Key Takeaways
An open-source tool can analyze customer sentiment and topics from call recordings.
Key technologies used include Whisper (speech-to-text), BERTopic (topic modeling), and Streamlit (UI).
This enables businesses to extract valuable insights from unstructured voice data.
Why it matters: This provides an accessible way for Indian tech companies to enhance customer service and product development by systematically understanding customer feedback.
MIT has recognized Jacob Andreas (EECS) and Brett McGuire (Chemistry) with the prestigious Edgerton Award. This accolade celebrates their outstanding achievements in teaching, groundbreaking research, and dedicated service to the MIT community. Their contributions underscore the interdisciplinary nature of AI advancement at one of the world's leading tech institutions.
Key Takeaways
MIT's Edgerton Award honours exceptional faculty contributions.
Jacob Andreas (EECS) and Brett McGuire (Chemistry) are the 2023 recipients.
The award recognizes excellence across teaching, research, and service.
Why it matters: This award highlights the crucial role of faculty in driving innovation and excellence in AI research and education at leading global universities.
This Towards Data Science article delves into advanced, often un-tutorialized aspects of building Large Language Models (LLMs) from the ground up. It covers critical optimization techniques like rank-stabilized scaling and quantization stability, offering a deep statistical and architectural perspective on what powers modern Transformer models. The author shares practical, hard-won insights gained through hands-on LLM construction, going beyond theoretical explanations found in typical guides.
Key Takeaways
Rank-stabilized scaling is a crucial but often overlooked optimization for LLM stability.
Achieving quantization stability requires careful statistical and architectural considerations.
Building LLMs involves practical, nuanced challenges not adequately covered in beginner tutorials.
Why it matters: Understanding these advanced, practical optimizations is key for Indian tech professionals looking to build or fine-tune LLMs effectively for local use cases and to contribute to the growing AI ecosystem.
KDnuggets highlights five essential Python scripts designed to automate advanced data validation and quality checks, crucial for robust modern data pipelines in India. These scripts address common data integrity issues like missing values and schema inconsistencies, offering smart solutions for developers and data scientists. By leveraging these tools, professionals can ensure cleaner, more reliable datasets for improved analysis and AI model performance.
Key Takeaways
Automate detection of missing values in datasets using Python.
Implement schema validation to catch structural data errors.
Enhance data quality through programmatic checks before analysis or deployment.
Utilize practical Python scripts for efficient data integrity assurance.
Why it matters: Ensuring data quality is paramount for the success of any data-driven initiative, especially in India's rapidly growing tech and AI landscape.
This Towards Data Science article delves into the critical aspect of memory for autonomous LLM agents, moving beyond theoretical concepts to offer practical guidance. It explores various architectures, common pitfalls, and effective patterns that enable these agents to retain and leverage information effectively. For Indian tech professionals, understanding these memory mechanisms is key to building more sophisticated and context-aware AI applications.
Key Takeaways
Effective memory is crucial for autonomous LLM agents to maintain state and perform complex tasks.
The article covers different memory architectures, highlighting their pros and cons.
It provides actionable insights into avoiding common memory-related challenges and implementing successful patterns.
Why it matters: Mastering memory for LLM agents is foundational for developing truly intelligent and persistent AI systems, impacting everything from chatbots to advanced robotics.
#LLM#Autonomous Agents#AI Memory#Towards Data Science#India Tech
The Cloud Native Computing Foundation (CNCF) is sounding the alarm: relying solely on Kubernetes for securing Large Language Model (LLM) deployments is a risky oversight. While Kubernetes is adept at workload orchestration and isolation, it lacks inherent understanding of AI system behaviors. This creates a unique and complex threat landscape that standard Kubernetes security practices alone cannot address.
Key Takeaways
Kubernetes's strength in workload management doesn't translate to inherent AI security.
LLM deployments introduce a novel and complex threat model beyond traditional infrastructure security.
Organizations need to adopt AI-specific security measures in addition to Kubernetes best practices.
Why it matters: As AI integration deepens, understanding and mitigating its unique security vulnerabilities is paramount for robust cloud-native deployments.
Anthropic has rolled out an agent-based pull request review system for its Claude Code, effectively introducing AI-powered code review capabilities. This new feature leverages multiple AI reviewers to analyze code changes, aiming to enhance the efficiency and quality of code reviews. The system is designed to provide comprehensive feedback on submitted code, assisting developers in identifying potential issues and improving code maintainability.
Key Takeaways
Anthropic's Claude Code now offers an agent-based pull request review system.
Multiple AI reviewers are employed to analyze code changes.
The feature aims to enhance code review efficiency and quality.
Why it matters: This advancement signifies a step towards more sophisticated AI integration in the software development lifecycle, potentially accelerating development cycles and improving code reliability in the Indian tech landscape.
Lakehouse architectures, while lauded for enabling diverse query engines to access shared data via formats like Apache Iceberg, are facing an interop challenge. The core issue lies in inconsistencies across database engines regarding SQL identifier resolution and catalog naming conventions. This can lead to unexpected failures when data is accessed by different engines within the same lakehouse, hindering true cross-engine compatibility.
Key Takeaways
Lakehouses using open table formats like Iceberg face identifier resolution conflicts between different SQL engines.
Discrepancies in catalog naming and SQL identifier rules are the root cause of these interoperability failures.
Enforcing consistent naming conventions and cross-engine validation is crucial for seamless lakehouse operations.
Why it matters: Resolving these identifier conflicts is essential for realizing the full potential of lakehouse architectures, ensuring data teams can efficiently leverage multiple tools on a single data foundation.
AWS has unveiled Agent Registry in preview, a new feature within Amazon Bedrock AgentCore designed to tackle the growing challenge of unmanaged AI agents within enterprises. This centralized catalog will allow organizations to discover, govern, and reuse AI agents, tools, and MCP servers, regardless of their deployment location. It natively supports both MCP and A2A protocols, offering a potential solution to AI sprawl and an alternative to offerings from Microsoft, Google Cloud, and ACP Registry.
Key Takeaways
AWS Agent Registry aims to centralize AI agent management and governance.
It provides a unified catalog for discovering and reusing AI agents and tools across an organization.
The preview feature supports multiple deployment environments and key protocols like MCP and A2A.
This move addresses the growing concern of 'AI agent sprawl' in enterprise settings.
Why it matters: For Indian tech-savvy readers, this launch signifies AWS's proactive approach to enterprise AI adoption challenges, potentially streamlining AI development and deployment within businesses as they increasingly integrate AI solutions.
The Daily AI Digest is an automated curation of the top 30 artificial intelligence news stories published across the web, summarized for quick reading.
How are these news articles selected?
Our system scans over 50 leading AI research labs, tech publications, and developer forums, evaluating factors like source authority, topic relevance, and community engagement to select the most important stories.
How often is the daily page updated?
The daily page is automatically generated every morning, ensuring you wake up to the most critical developments from the previous 24 hours.
What sources do you track for AI news?
We track a diverse range of sources, including mainstream tech media (like TechCrunch), AI-specific publications (like The Batch), academic institutions (Stanford HAI), and major lab blogs (OpenAI, DeepMind).
How does the AI summarize the articles?
We use advanced large language models (currently Gemini) to process the content of the selected articles and extract the core narrative, key takeaways, and broader significance.
Can I see news from previous days?
Yes, you can navigate to previous dates using the date navigation at the top of the page, or browse the complete chronological archive.
How do you decide which news is most important?
Importance is judged by a combination of algorithmic analysis separating signal from noise, and manual weighting of authoritative sources over aggregate sites.
Are the AI summaries reliable?
While highly accurate, AI summaries are generated representations of the source material. We always provide a 'Read Original' link so you can verify facts directly with the primary source.
Do you include research papers in the daily news?
Yes, major breakthroughs published on platforms like Papers With Code or arXiv are picked up if they generate significant academic or industry buzz.
Can I get these updates via email?
Currently, the digest is web-only, but an email newsletter feature is on our roadmap for future development.