If search-agent accuracy has plateaued, the next thing to cut is context tokens. Context as Infrastructure: Knowledge Compilation for Retrieval in Agentic Systems with Harnesses A harness using only BM25 and grep reportedly reaches 91% on FRAMES. What's left is wasted tokens from rereading the same documents. The paper proposes "knowledge compilation": adding context offline with an LLM. 🧾 Highlight 1: claim-evidence pairs Each unit is a normalized claim, a verbatim evidence span, and a source offset. Claims without a locatable span are not stored. When retrieval finds a chunk, the chunk is replaced by these pairs before it enters the context window. 📉 Highlight 2: structure didn't help, substitution did Summary trees, outlines, wikis and graphs showed no detectable accuracy gain. Claim-evidence substitution cut query tokens by 38% on average on FRAMES (N=200) with accuracy within noise. The cost is more retrieval rounds and higher latency, which a decision gate helps contain. 🧮 Highlight 3: a payback formula With gate acceptance rate p_h, break-even is about 2%. The measured average implies roughly 43%, so a compile step costing 10 queries pays back after about 26 queries. Caveats: one benchmark, single runs, no confidence intervals, and the false-accept audit hasn't been run. Treating context as infrastructure feels practical to me. #RAG# #AgenticAI#

From mountains and hills to plateaus and plains, land is life. It provides oxygen and acts as an essential carbon sink that regulates the planet's temperature and absorbs its carbon emissions. Yet unsustainable use is putting these vital ecosystems at risk. See what can be done:

'Aerial herders' take off on Qinghai's plateau pastures At the foot of snow-capped mountains, on the vast grassland, a drone is on a herding mission. In Domkog township, Maqen county, Golog Tibetan Autonomous Prefecture, northwest China's Qinghai Province, herdsman Gongpo Tsering began using a drone to herd livestock this year. With an aerial view, he can quickly track the movements of his yak herd. The buzzing sound of the drone also helps drive the yaks down from the mountains. Once dependent on horseback to patrol the pastures, he now relies on technology to guard his livestock, making his work and daily life more efficient and convenient.

Demand for energy in the US plateaued after 2000. It’s over determined, but perhaps this was because we didn’t have a good use for more energy. AI training/inference means there is always a marginal use for more power. The marginal megawatt is now more intelligence

Starlink is helping strengthen child protection in Nigeria 🇳🇬 Plateau State has received 40 Starlink kits for child care institutions, each with 13 months of sponsored internet service The connectivity will support child welfare records, case management, family tracing, and coordination between care homes and government agencies Poor internet has been slowing updates and case reviews Now Starlink is helping these institutions stay connected and keep critical information moving