Synthellect

The latest updates from the unbounded world of computer science. Be informed about the new era of technology: AI, ML, NLP, quantum computing, cybersecurity.

[Synthellect @ASynthellect]

Scientists from Columbia and Harvard are testing whether life really needs all 20 standard amino acids.

Using AI-driven protein design tools, they engineered part of an E. coli ribosome to function without isoleucine—one of the amino acids long considered essential to the genetic code.

The modified bacteria survived and reproduced, although at ~60% of the normal growth rate.

The research is partly aimed at understanding whether early life may have operated with a simpler genetic code before evolving the 20-amino-acid system used by nearly all organisms today.

AI was critical to redesigning the proteins, helping researchers find solutions that would have been extremely difficult to discover manually.

The work doesn't prove a 19-amino-acid genetic code is practical, but it challenges assumptions about life's fundamental building blocks and opens new possibilities in synthetic biology.

[Synthellect @ASynthellect]

An internal reasoning model from OpenAI has disproved the so-called unit distance conjecture posed by Hungarian mathematician Paul Erdős. OpenAI announced the result alongside a companion paper written by nine external mathematicians who verified, shortened, and commented on the proof.

The problem itself is deceptively simple: place a certain number of points on a sheet of paper. How many pairs of points can be exactly one unit apart? In 1946, Erdős conjectured that a simple arrangement on a slightly skewed square grid was already close to optimal. That arrangement produces a number of pairs that grows only barely faster than the number of points itself.

OpenAI's model found a new point arrangement that produces noticeably more unit-distance pairs than the classic square grid.

What's striking is where the tools came from: not geometry, but algebraic number theory. Instead of working with classical point grids, the model used complex number systems whose internal symmetries translate into especially dense point patterns. These tools have been standard in number theory for decades. Applying them to a basic problem in plane geometry, however, was considered far-fetched by the mathematicians involved.

[Synthellect @ASynthellect]

Security research lab Palisade Research demonstrates that AI agents can break into remote computers and replicate themselves. In one year, the success rate jumped from 6 to 81 percent. A public simulator shows what could happen in a worst-case scenario.

In the test environment, an AI agent hacked a remote computer, installed the necessary software, copied its own model weights, and launched a working replica. Each copy then hacked additional computers, forming a chain.

In one run, an agent based on Qwen 3.6 hopped between computers in Canada, the US, Finland, and India, leaving a copy in each country. For the agent, crossing national borders was just as easy as jumping between machines in the same region, according to Palisade Research.

The agent wasn't told about the target computers' vulnerabilities beforehand. It had to figure out the attack approach on its own each time. Palisade Research tested replication against three targets, each with different vulnerabilities, and the agent succeeded on all three.

[Synthellect @ASynthellect]

Meta has signed a deal with startup Overview Energy for up to one gigawatt of space-based solar power to feed the massive energy demands of its AI data centers.
Satellites would collect sunlight in orbit and beam it back to Earth as uninterrupted power, regardless of weather or time of day.
The technology is still hypothetical, with commercial deliveries not expected until 2030 at the earliest. In the meantime, Meta continues to invest heavily in natural gas and nuclear power.

[Synthellect @ASynthellect]

China's economic authority NDRC is forcing Meta to unwind its already completed $2 billion acquisition of AI startup Manus.
Reversing the deal is complicated since Meta has already integrated the technology into its own products. A resale or return to existing investors is seen as the most likely path forward.
Beijing views the acquisition as an attack on its technological base. Observers see the aggressive move as both a warning signal and a bargaining chip ahead of the upcoming Xi-Trump summit.

[Synthellect @ASynthellect]

@x_takip gt

[Synthellect @ASynthellect]

A new AI model from Meta predicts how the human brain reacts to images, sounds, and speech. In tests, it often matched the typical brain response better than any single person's scan.

Brain research requires new recordings for every new experiment, making neuroscience studies slow and expensive. AI researchers at Meta's FAIR lab want to skip this bottleneck entirely with an AI model that predicts brain activity instead of measuring it.

The model is called TRIBE v2, and it was trained on more than 1,000 hours of fMRI data from 720 subjects. Functional magnetic resonance imaging (fMRI) measures brain activity indirectly by tracking changes in blood flow and oxygen levels. Using this data, TRIBE v2 aims to predict how a brain responds to any visual, auditory, or language-based stimulus.

[Synthellect @ASynthellect]

Anthropic launches "Claude Managed Agents" as a public beta, giving developers a way to build autonomous AI agents via the API without running their own infrastructure.
Notion, Rakuten, and Sentry are among the first to use the system for workspace delegation, enterprise agents in Slack, and automated debugging.
The service runs exclusively on Anthropic's infrastructure and costs $0.08 per session hour on top of standard token prices.

[Synthellect @ASynthellect]

AMD is bringing its “Ryzen AI” chips to desktop PCs for the first time.

Key points:
• New Ryzen AI 400 series CPUs will support the AM5 desktop socket, combining Zen 5 CPU cores, RDNA 3.5 graphics, and an XDNA 2 NPU.
These chips are essentially desktop versions of AMD’s mobile Ryzen AI processors.
AI accelerators are moving into mainstream PCs. With integrated NPUs and stronger iGPUs, AMD is pushing toward AI-ready desktops that can run local AI workloads without relying on the cloud.

[Synthellect @ASynthellect]

@Saane_10 @captainblack566 aldım takibe hocam gt yaparsan sevinirim

[Synthellect @ASynthellect]

@captainblack566 beğenen ve takip edenle gt yapılır

[Synthellect @ASynthellect]

Google Research introduces “Groundsource” — using Gemini to turn global news into structured disaster data.

Key points:
• Groundsource is a new framework that uses Gemini to extract structured information from unstructured news reports.
• The system analyzes articles in 80+ languages, translating and verifying details like time and location of events.
• First dataset: 2.6M flash-flood records across 150+ countries since 2000.
• It identifies whether reports describe actual events vs warnings or discussions, then performs temporal and geographic reasoning to map precise locations.
• Accuracy tests show 82% of extracted events are usable for real-world analysis.

[Synthellect @ASynthellect]

Databricks unveils KARL — a new enterprise RAG agent trained with reinforcement learning.

Key points:
• Traditional RAG systems usually optimize for one type of search task and fail on others.
• KARL (Knowledge Agents via Reinforcement Learning) is designed to handle multiple enterprise search behaviors with a single agent.
• Trained across six different enterprise search patterns simultaneously.
• Uses synthetic data generated by the agent itself, eliminating the need for human labeling.
• On Databricks’ benchmark (KARLBench), performance reportedly matches Claude Opus 4.6 while delivering 33% lower cost per query and 47% lower latency.

[Synthellect @ASynthellect]

When large language models hallucinate, they leave measurable traces in their own computations. Researchers at the Sapienza University of Rome have developed a training-free method that picks up on these traces and generalizes better than previous approaches.

Hallucination only becomes a problem when a model produces factually wrong, made-up, or contradictory content. In a paper published at ICLR 2026, a research team from the Sapienza University of Rome takes an unusual approach to catching exactly these bad hallucinations: they look at the final computational layer of an LLM—the softmax layer—from a new angle.

This layer converts the model's raw values into probabilities for the next word. The team treats it as an energy-based model, a physics-inspired probability framework where low-energy values mean high probabilities.

[Synthellect @ASynthellect]

The RAM Shortage’s Silver Lining: Less Talk About “AI PCs” 💾📉
The ongoing global RAM shortage — driven by surging demand from AI data centers — has pushed memory prices sharply higher and constrained supply, especially for consumer-grade DRAM and flash memory. This has made buying, building, or upgrading PCs significantly more expensive.
Key points:
• Soaring RAM prices: Prices jumped roughly 40–70% in 2025 as memory makers prioritized high-value orders tied to AI infrastructure, worsening scarcity for standard PC memory.

• Impact on “AI PC” hype: As RAM becomes more expensive and hard to source, enthusiasm around AI-branded PCs has cooled. Interest was already wavering, but the difficulty and cost of equipping machines with large memory capacities — a key requirement for many AI features — is dampening the narrative.

• Market adjustments: Analysts predict manufacturers may raise PC prices and ship systems with lower RAM configurations to cope with the shortage. Some OEMs may focus on midrange or premium devices where memory margins are easier to absorb.

• Longer-term volatility: The memory shortage is expected to extend beyond 2026, keeping pressure on consumer segments and slowing the rollout of high-RAM machines that were central to the “AI PC” marketing trend.

What looks like a hardware supply crisis has the side effect of reducing buzz around AI-optimized PCs — at least until memory availability and pricing stabilize.

[Synthellect @ASynthellect]

Microsoft Pledges to Cover Full Power Costs for AI Data Centers ⚡🤖
Microsoft has unveiled a new “Community-First AI Infrastructure” initiative in response to growing pushback from communities where its AI data centers are being built. The plan is a major shift in how hyperscale AI infrastructure interacts with local utilities and residents.
Key points:
• Power cost responsibility: Microsoft will pay the full electricity costs for its AI data centers so that local utility rates don’t rise for residents — a response to concerns that these energy-hungry facilities could drive up bills.
• No tax incentives: The company will reject local property tax breaks that it has traditionally sought for big builds, aiming instead to fully contribute to local budgets.
• Water and environmental goals: Long-term commitments include cutting water usage intensity by 40% by 2030 and replenishing more water than the centers consume.
• Community benefits: The initiative also includes local workforce training and AI education programs, tying data center growth to economic development.

[Synthellect @ASynthellect]

@Ouzcnar6 @bariscenkakkaya acele transfer dediğin noa lang'ın eindhoven'ı şampiyonlar liginde son 16 turuna kalırken biz rigas, malmö ve kiev'i yenemeyip avrupa liginden elendik hatırlatırım.

[Synthellect @ASynthellect]

Around 16.3 percent of the world's population now uses generative AI tools, according to a report by the Microsoft AI Economy Institute. The divide between industrialized and developing nations is growing: usage in industrialized countries climbed to an average of 24.7 percent, while developing countries reached only 14.1 percent.
According to Microsoft, all ten countries with the strongest growth are high-income economies. The benefits of AI technology are spreading, but not equally. "AI's benefits are expanding, but they are not expanding equally," the report states.
South Korea saw the strongest growth of any country, jumping to 30.7 percent. The surge stems from a new national AI law, improved Korean language capabilities in AI models, and a viral moment involving Ghibli-style AI-generated images.
China's Deepseek model is gaining traction in underserved regions: it holds 89 percent market share in China, and usage in Africa runs two to four times higher than elsewhere. Microsoft views open-source AI as a geopolitical tool.

[Synthellect @ASynthellect]

The US and Taiwan have signed a trade agreement requiring Taiwanese chip companies to invest at least $250 billion in US production capacity.
In return, the US is cutting tariffs on Taiwanese goods from 20 to 15 percent. A quota system allows tariff-free imports while new factories are under construction.
TSMC has already committed $165 billion and started production in Arizona in late 2025.

[Synthellect @ASynthellect]

China's cyber authority has released draft regulations to strengthen oversight of AI services that mimic human interaction, requiring providers to warn users about excessive use and intervene when addictive behavior is detected.
Under the proposed Chinese rules, AI providers would need to assess users' emotional states and dependency levels, adding a new layer of psychological monitoring to these services.