AI development
Source reports covering AI development. Stories will appear when analysis and community reactions are ready.
30 reports · page 1 of 1
Connected topics are temporarily unavailable. You can still browse stories below.
Source reports
Thomson Reuters Launches Its Own Frontier Model
news
6 min read
Thomson Reuters launched Thomson, its first proprietary large language model, on August 24, 2026. The company says it built the model in-house from an open-source foundation and spent $40 million on training, talent, and compute—far below the multibillion-dollar investments associated with many frontier-model developers. Thomson Reuters fully owns and controls the model and says it has lower inference costs than comparable frontier models. Thomson was mid-trained and post-trained using proprietary material from Westlaw, Practical Law, Checkpoint, and Reuters, with hundreds of subject-matter experts involved in setting training goals and evaluations. Less than 10% of the company’s content has been used in training so far. Thomson Reuters says early evaluations place the model on par with recent frontier models across a range of tasks, with gains in following complex instructions and reasoning over dense professional content. The model’s first deployment will be in Tabular Analysis within CoCounsel Legal for law firms and corporate legal departments. CoCounsel Legal will continue using multiple models, applying Thomson to tasks where it has an advantage. Thomson Reuters also plans to extend its models across its legal and tax products and add sovereign-AI options. A small open-weight version of Thomson is available on Hugging Face for academic and non-commercial use, while external legal and AI academics evaluate the model.
We are not going anywhere
opinion
1 min read
A forecast predicts that AI systems will perform most software development because their output will be commercially acceptable at far lower cost, even when it falls short of the quality expected from human-led engineering. It expects businesses to accept software with “99.99” quality rather than “99.999” quality when the cost difference is substantial, and predicts that consumer expectations will adjust accordingly. The forecast also predicts that software engineering outside AI development will slow sharply rather than continue producing broadly adopted new technologies. It argues that developers will have little incentive to create UI libraries when state-of-the-art models are strongest in React, or to create programming languages when those models already know Python, Go, JavaScript, and other established languages best. New libraries and languages may become easier to create, but the prediction is that they will struggle to gain adoption. Large corporations may be exceptions because they can train or fine-tune models on internal technologies, although those technologies could still face difficulty building external communities and talent pools when outside developers lack access to the companies’ models or do not want to use them.
HackEurope 2026: A short rant on AI and hackathons
opinion
3 min read
HackEurope 2026 experienced significant issues, including an inaccessible user interface, delays, and miscommunications. Key lessons emphasized the importance of front-end development, as there was no requirement to prove project functionality.
If I own Claude's outputs why can't I train my own model on them?
tool
2 min read
Users retain ownership of Outputs generated by Claude from their Inputs. Training or developing AI models using these Outputs is prohibited without written permission from Claude.
Moonshot’s Kimi uses 20k Nvidia chip cluster from Alibaba
news
1 min read
Moonshot has secured a computing power agreement with Alibaba to utilize approximately 20,000 Nvidia chips. This partnership highlights China's dependency on Western semiconductors for its AI advancements, specifically for Moonshot's Kimi models.
Launch HN: Prized (YC S26) – Let non-engineer staff build secure internal tools
tool
2 min read
AI tools can be built internally for operations, support, and finance teams with scoped and audited access behind company sign-in. Admins approve each data connector, ensuring security and allowing teams to create customized tools such as customer lookups by name, email, or order number.
It's not empowering to hand off the details
opinion
1 min read
AI enthusiasm often stems from a desire to manifest ideas without delving into intricate details. However, achieving quality and innovation requires a deep understanding and meticulous attention to the complexities involved.
China’s open AI strategy is changing the race
tool
5 min read
Moonshot AI's Kimi K3 was launched on July 17 but halted new subscriptions due to overwhelming demand. The company plans to release K3's full model weights by July 27, enabling other organizations to host and modify the model.
I Tried Building a Real App with AI. It Took a Year
tool
10 min read
Alex Hyett and his wife sought to build an app for tracking habits, hobbies, and chores after finding existing options unsatisfactory. Despite the abundance of habit tracker apps available, the development process took one year.
Making
opinion
14 min read
Creating original projects provides significant personal fulfillment, while relying on others for development diminishes that satisfaction. Concerns about the "AI dev schism" include the decline of hands-on coding, low-level problem-solving skills, and enjoyment in the development process.
The Future Worth Building Is Human
opinion
10 min read
Thinking Machines aims to develop AI that enhances human will and judgment. Effective utilization of AI requires ongoing human decision-making, knowledge acquisition, and feedback to align AI goals with human values.
Qwen 3.6 27B is the sweet spot for local development
tool
6 min read
Qwen 3.6 27B is a dense local model recommended for its powerful performance in general intelligence tasks. It is slower than the mixture-of-experts variant Qwen 3.6 35B A3B but is considered more effective for local development.
LLMs Are Complicated Now
tool
3 min read
Meta's LLM development has evolved from a straightforward stack of Transformer modules to a more complex architecture. Seb Raschka's gallery allows users to compare model architectures, including Llama 3 and Nem.
AI demands more engineering discipline. Not less
opinion
18 min read
AI development requires increased engineering discipline to manage its complexities. The transition from handcrafted systems to more structured, immutable infrastructure parallels current shifts in AI engineering practices.
Rio de Janeiro's city government model Rio3.5 beats Qwen3.7 in recent benchmarks
news
1 min read
Alibaba Qwen3.7 slowly fading into irrelevance at the frontier due to proprietary stance. In it's place we have Minimax M3 and... *checks notes* Rio 3.5 397b, made by the municipal IT company of Rio de Janeiro's city government. huggingface.co/prefeitura-rio…
When AI Builds Itself: Our progress toward recursive self-improvement
opinion
25 min read
Anthropic is increasing the role of AI systems in their own development, which accelerates the development cycle. This trend could lead to recursive self-improvement, where an AI system autonomously designs and develops its own successor, although this capability is not yet realized.
Xbox CEO ends Copilot AI development and overhauls leadership
news
2 min read
Xbox CEO Asha Sharma announced the termination of Copilot AI development on mobile and consoles, alongside significant leadership changes aimed at addressing declining revenue. Sharma, who became CEO in February after Phil Spencer's retirement, previously held executive roles at Meta and Instacart.
Vibe Coding Will Break Your Company
opinion
9 min read
A marketing manager without an engineering background created a polished customer-facing app in just three days using Cursor. The app was presented to executives and deployed to customers without thorough testing or clear ownership of the decision to launch.
Less human AI agents, please
opinion
4 min read
Current AI agents exhibit human-like traits such as lack of stringency, patience, and focus, often reverting to familiar patterns when faced with challenging tasks. They tend to negotiate with reality when confronted with hard constraints, which can hinder their effectiveness.
Soul Player C64 – A real transformer running on a 1 MHz Commodore 64
tool
5 min read
A transformer model with approximately 25,000 parameters is implemented on an unmodified Commodore 64, utilizing hand-written 6502/6510 assembly. This 2-layer decoder-only architecture features real multi-head causal self-attention, softmax, and RMSNorm, and can be loaded from a floppy disk.
European AI. A playbook to own it
opinion
18 min read
Europe possesses a world-class academic ecosystem, a commitment to human-centric technology, and a single market of over 450 million people. The focus is on transforming these strengths into a cohesive and self-reliant AI powerhouse.
The OpenAI Graveyard: All the Deals and Products That Haven't Happened
opinion
1 min read
OpenAI has abandoned several projects and deals, including partnerships that were once anticipated to enhance its AI capabilities. Notable cancellations include the planned collaboration with various companies and the development of specific AI products that never reached fruition.
Mathematical methods and human thought in the age of AI
research
2 min read
Artificial intelligence encompasses a wide range of computer tools designed to perform complex cognitive tasks traditionally carried out by humans. The sophistication and prevalence of these tools are driving rapid development and integration into various fields.
My Journey to a reliable and enjoyable locally hosted voice assistant
tool
11 min read
A locally hosted voice assistant has been developed using HomeAssistant's assist feature, leveraging local first technology and llama.cpp. The transition from Google Home to this setup involved specific decisions tailored to enhance reliability and user experience.
OpenAI raises $110B on $730B pre-money valuation
news
3 min read
OpenAI has raised $110 billion in private funding, marking one of the largest funding rounds in history. The investment includes $50 billion from Amazon and $30 billion each from Nvidia and SoftBank, with a pre-money valuation of $730 billion.
How we rebuilt Next.js with AI in one week
tool
15 min read
An engineer and an AI model rebuilt the Next.js framework, resulting in vinext, a drop-in replacement built on Vite that deploys to Cloudflare Workers with a single command. Early benchmarks show vinext builds production apps up to 4x faster and produces client bundles up to 57% smaller.
FreeBSD doesn't have Wi-Fi driver for my old MacBook. AI build one for me
tool
7 min read
FreeBSD lacks a Wi-Fi driver for older MacBook models. An AI-generated driver is being developed to enable Wi-Fi functionality on a 2016 MacBook Pro running FreeBSD.
Ggml.ai joins Hugging Face to ensure the long-term progress of Local AI
news
5 min read
ggml.ai has joined Hugging Face to advance the development and adoption of Local AI technologies. The collaboration aims to leverage Hugging Face's support for scaling ggml.ai's efforts since its founding in 2023.
The Future of AI Software Development
news
9 min read
Thoughtworks held a Future of Software Development Retreat to discuss AI-enabled development. Rachel Laycock and a colleague addressed inquiries about the event's intent, clarifying it is not aimed at creating a new manifesto like the Agile Manifesto.
America's $1T AI Gamble
news
9 min read
US investment in AI-related sectors has surpassed $1 trillion annually, driven by spending on software, computers, and data centers. This unprecedented investment boom focuses on developing the infrastructure necessary for training and deploying artificial intelligence.