Anthropic settlement copyright lawsuit reports trend, OpenAI and Google strike gold at ICPC

AI_Distilled #113: What’s New in AI This Week

The past week has brought both reckoning and recognition for the AI industry. Anthropic’s billion-dollar copyright settlement is making publishers and creators wonder how far accountability will go, even as the company reaches a confidential deal with The New York Times. Meanwhile, the competitive spirit is alive and well: OpenAI and Google claimed gold-standard performances at the ICPC World Finals, cementing their models’ ability to tackle some of the toughest programming challenges on the planet. And as regulators circle, TikTok’s U.S. sale edges closer to completion.

LLM Expert Insights,

Packt

📈LATEST DEVELOPMENT

OpenAI strikes gold at ICPC, releases GPT-5Codex, and finds it harder to train scheming models

OpenAI achieved gold-medal level performance at the 2025 ICPC World Finals, solving all the 12 problems, at the finals in Baku, Azerbaijan (ICPC Finals).

GPT-5-Codex, optimized for real-world software engineering: quicker interactions, better at long tasks, refactoring, complex code review. Codex now integrates seamlessly via CLI, IDE, cloud, GitHub, promising more reliable, context-aware coding. Check it out here!

OpenAI, with Apollo Research, found frontier models showing “scheming” (pretending alignment while pursuing hidden agendas). To tackle this the team has introduced deliberative alignment training, reducing covert actions ~30× in tests. You can read more about it here.

Google DeepMind’s AI breakthroughs for the future of AGI and fluid dynamics

Gemini 2.5 Deep Think too achieved gold-medal level performance at the 2025 ICPC World Finals, solving 10 of 12 problems under competition conditions. Read more here!

In another development reported by Google, DeepMind has developed new AI methods using Physics-Informed Neural Networks (PINNs) to discover previously unknown unstable singularities in major fluid dynamics equations. These findings advance mathematical understanding of “blow-ups” in systems like the Navier-Stokes and Euler equations. Read more here!

From MongoDB's Director of Engineering: Build AI-Powered Platforms

anthropic-settlement-copyright-lawsuit-reports-trend-openai-and-google-strike-gold-at-icpc-img-1

Last Few Tickets Left: 30% off for 72 Hours

EXPERT INSIGHTS

Mastering Text Data Augmentation Techniques for LLMs

Text data augmentation is a foundational technique for enhancing the robustness and generalization of large language models (LLMs). By artificially increasing the size and variability of training data, models can be exposed to more diverse linguistic patterns, reducing overfitting and improving their ability to handle unseen inputs. This tutorial walks through key augmentation techniques that are particularly effective when working with text-based LLMs.

Synonym replacement: This is a classic approach that involves replacing words with their synonyms to produce semantically equivalent sentences. Using lexical databases like WordNet, each word in a sentence is checked for synonyms, which are then substituted randomly. This introduces variation without altering the meaning drastically, while ensuring syntactic and semantic consistency, particularly useful for simpler augmentation tasks. The code below replaces words in a sentence with synonyms from WordNet to create varied yet semantically consistent text:

from nltk.corpus import wordnet
												import random
												
												def synonym_replacement(text):
												 words = text.split()
												 new_words = []
												 for word in words:
												 synonyms = wordnet.synsets(word)
												 if synonyms:
												 synonym_words = [lemma.name() for s in synonyms for lemma in s.lemmas() if lemma.name() != word]
												 if synonym_words:
												 new_words.append(random.choice(synonym_words))
												 else:
												 new_words.append(word)
												 else:
												 new_words.append(word)
												 return ' '.join(new_words)
												
												print(synonym_replacement("The quick brown fox jumps over the lazy dog."))

Read the full article on Substack →

Built something cool? Tell us.

Whether it's a scrappy prototype or a production-grade agent, we want to hear how you're putting generative AI to work. Drop us your story at nimishad@packtpub.com or reply to this email, and you could get featured in an upcoming issue of AI_Distilled.

📢 If your company is interested in reaching an audience of developers and, technical professionals, and decision makers, you may want toadvertise with us.

If you have any comments or feedback, just reply back to this email.

Thanks for reading and have a great day!