Sign in

Elizabeth Fuentes L

@elizabethfue12.bsky.social
153 followers 80 following 925 posts

AWS Developer Advocate 🇻🇪🇨🇱. I work for @AWSCloud & my opinions are my own. Soy desierto, selva, nieve y volcán…

PostsRepliesMedia
Elizabeth Fuentes L @elizabethfue12.bsky.social · 06/10/2026
DynamoDB: 192 million requests/second during Prime Day 2026. Single-digit millisecond responses. Lambda: 2.3 trillion invocations/day. Production scale, not benchmarks. #AWS #DistributedSystems ift.tt/Lc5tKEb
000
Elizabeth Fuentes L @elizabethfue12.bsky.social · 05/10/2026
Amazon Bedrock agents now run OpenAI models with AWS-native execution. Self-hosted or managed runtime, your permissions. GPT-6.1 Sol: 1/5 the cost of Astra for agentic work. Astra UltraFast: 300 tokens/sec. 🔥 #GenerativeAI #AWSBedrock ift.tt/UjdCc40
010
Elizabeth Fuentes L @elizabethfue12.bsky.social · 02/10/2026
Decision models choose from options instead of generating text. The tradeoff: faster, more capable at a given size, built-in reliability scores. AWS open-sourced Strands Decider 2B. 2B params, runs locally, answers in milliseconds. Training data include…
ift.tt
Introducing Strands Decider 2B: a small, open source, decision model
Strands Decider 2B is a small, open source, decision model optimized for fast experimentation, local development, and innovation.
021
Elizabeth Fuentes L @elizabethfue12.bsky.social · 01/10/2026
AWS Well-Architected Agent is now in preview. It runs architecture reviews automatically across 65+ services and generates recommendations with IaC changes and cost impact. Finally. #AWS #CloudArchitecture ift.tt/Xkwgsuo
020
Elizabeth Fuentes L @elizabethfue12.bsky.social · 01/10/2026
AWS launched 3 new Jam experiences for agentic AI: spec-driven dev with Kiro, building agents on Bedrock AgentCore, and multi-agent orchestration. Hands-on challenges, real services, actual production patterns. #AgenticAI #AWSBedrock
ift.tt
Build, deploy, and harden AI agents: Three new AWS Jam experiences now available  | Amazon Web Services
You write specs that guide AI coding agents, deploy agents on managed runtimes, and orchestrate multi-agent systems across production workloads. Now you can build those skills in a live AWS environment. ...
132
Elizabeth Fuentes L @elizabethfue12.bsky.social · 30/09/2026
S3 Tables now support all Iceberg V3 data types. Deletion vectors replace positional delete files. Deleting 50k rows from a 2B-row table writes one file instead of thousands. Native variant, nanosecond timestamps, geometry types. #ApacheIceberg #DataEngineering ift.tt/Jxf1Ahu
000
Elizabeth Fuentes L @elizabethfue12.bsky.social · 30/09/2026
S3 Vectors now filters metadata before the similarity search, so you get higher recall on narrow queries. Each vector carries 2 KB of filterable metadata. No re-ingestion, no cost. #VectorSearch #AWS ift.tt/IJKT9Rh
010
Elizabeth Fuentes L @elizabethfue12.bsky.social · 30/09/2026
Aurora PostgreSQL now queries Iceberg and Parquet data in S3 directly. No ETL, no duplication. DuckDB embedded in Aurora makes this possible. Available on PostgreSQL 17.11+ and 18.6+. #PostgreSQL #DataLakes ift.tt/hkMOc4d
000
Elizabeth Fuentes L @elizabethfue12.bsky.social · 30/09/2026
Three new AWS Heroes announced today. Avinash Dalvi: 40+ community talks on serverless Joanne Skiles: runs Transparent Systems lab at Rollins College Xiaofei Li: built 5,000+ member Chinese AWS community #AWSHeroes #CloudCommunity ift.tt/UuFiQlg
000
Elizabeth Fuentes L @elizabethfue12.bsky.social · 30/09/2026
AWS released 2 free hands-on assessments for agentic AI builders. You fix real identity and database issues in a live environment. No multiple choice. Pass = Credly badge. Both free on Skill Builder. #AgenticAI #AWS
ift.tt
Prove your agentic AI skills: Two new Microcredentials now available | Amazon Web Services
An AI agent is only as trustworthy as the identity controls and data layer behind it. If you’re the one building those, it’s time to validate — and showcase — exactly what you can do. Today, two...
010
Elizabeth Fuentes L @elizabethfue12.bsky.social · 28/09/2026
GPT-6 Sol and Luna + Claude Opus 5.5 are now on Amazon Bedrock. All three cheaper than their predecessors. The shift: you match the model to the job instead of reaching for the biggest one every time. #GenerativeAI #Bedrock ift.tt/UHF3M0s
011
Elizabeth Fuentes L @elizabethfue12.bsky.social · 24/09/2026
EventBridge now supports one shared event bus across all accounts in your AWS org. Native ordering, simplified subscribers, better pricing at scale. No more chaining buses with cross-account rules. #EventBridge #ServerlessAWS ift.tt/ZENmM60
000
Elizabeth Fuentes L @elizabethfue12.bsky.social · 22/09/2026
CloudWatch Omni is GA. You access it via a dedicated URL with SSO (no AWS Console needed). Your whole team shares the same investigation context, and it auto-discovers services from OpenTelemetry data. #CloudWatch #OpenTelemetry ift.tt/FqPTYrf
000
Elizabeth Fuentes L @elizabethfue12.bsky.social · 22/09/2026
CloudWatch Omni: trace every agent interaction with built-in evals for correctness, coherence, and tool selection. Works in your IDE, runs locally or in cloud, no console needed. #GenerativeAI #AIAgents ift.tt/Zf9JCmH
010
Elizabeth Fuentes L @elizabethfue12.bsky.social · 21/09/2026
AWS Builder Center now has iOS and Android apps. Browse articles, access 600+ courses, spin up workshop sandboxes from your phone. Your Builder ID syncs across sessions. 📱 #AWS #CloudComputing ift.tt/zbRtDcL
010
Elizabeth Fuentes L @elizabethfue12.bsky.social · 19/09/2026
New Tech & AI Weekly is out: the harness tax (same model, up to 5x cost), the best coding agent failing 60% on unseen code, GitHub's Rust rewrite of Copilot, AWS AgentCore tuning prompts from traces, plus my posts on caching and memory: blog.elifuentes.tech/blog/newslet...
001
Elizabeth Fuentes L @elizabethfue12.bsky.social · 17/09/2026
T8i instances are GA: 30% better price performance than T3, 70% more compute, 2.4x EBS bandwidth. Same burstable model, just swap the instance type. #AWS #EC2 ift.tt/dX02UwI
000
Elizabeth Fuentes L @elizabethfue12.bsky.social · 17/09/2026
Elastic Beanstalk Cluster Mode: deploy source code in 7 languages, AWS containerizes it automatically via Cloud Native Buildpacks and runs it on shared EKS infrastructure. No Dockerfile needed. Scales from 10 to 100 apps without adding ops complexity. #AWS #Kubernetes ift.tt/TIq8F9G
000
Elizabeth Fuentes L @elizabethfue12.bsky.social · 16/09/2026
Prompt injection can outlive the chat. If stored in memory, it can affect later sessions. This demo screens writes in the agent harness before storage. bit.ly/4h8lJ5g
dev.to
https://dev.to/aws/stop-ai-agent-memory-poisoning-at-the-write-path-1m9f
000
Elizabeth Fuentes L @elizabethfue12.bsky.social · 14/09/2026
GPT-6 Astra on Amazon Bedrock now supports 1M token context windows. Send entire codebases or document collections and let it reconcile competing inputs. Your inference data stays yours. 🔒 #Bedrock #GenAI ift.tt/LtuWPqR
000
Elizabeth Fuentes L @elizabethfue12.bsky.social · 12/09/2026
Prompt caching only discounts the input you resend, the model still generates every answer from scratch. New intro post + 4-part series on caching AI agents by meaning: response, reasoning & tool-result caches. 58% fewer tokens on a warm run. dev.to/aws/prompt-c...
111
Elizabeth Fuentes L @elizabethfue12.bsky.social · 11/09/2026
Sierra gave coding agents a company's docs, API and codebase and asked them to build a customer service agent. Best score: 23.9%. Nobody broke 25%. Then one sentence suggesting a different architecture took a build from 31% to 67%. blog.elifuentes.tech/blog/newslet...
000
Elizabeth Fuentes L @elizabethfue12.bsky.social · 10/09/2026
EBS volume clones now work across AWS accounts. Copy production volumes to dev/test accounts and re-encrypt with a different KMS key. #AWS #CloudComputing ift.tt/A67sh0b
000
Elizabeth Fuentes L @elizabethfue12.bsky.social · 07/09/2026
The agent that wins isn't the one that remembers the most. It's the one that keeps the right things and throws the rest away. I built memory extraction 3 ways in Strands to test that. Code and results in the repo. #AIAgents #GenerativeAI
ift.tt
AI Agent Memory: What to Store and What to Throw Away
Everyone is racing to make agents remember more. Bigger context windows, longer histories, a vector...
101
Elizabeth Fuentes L @elizabethfue12.bsky.social · 07/09/2026
Claude Fable 5.1 on AWS handles features across entire codebases and runs for hours. It's a Covered Model, AWS retains prompts for 30 days for safety review, but your data stays in the AWS boundary. #GenerativeAI #AWS ift.tt/5MYmzRC
000
Elizabeth Fuentes L @elizabethfue12.bsky.social · 02/09/2026
AWS updated the ML Engineer Associate cert to include generative AI, agentic AI, and LLM workflows. Beta (MLA-C02) open now. Current version ends Sept 28. New one tests real production ML + GenAI work. #MachineLearning #GenAI
ift.tt
Certification updates from AWS Training and Certification: September 2026 | Amazon Web Services
AWS Certification is updating three exams to reflect how cloud roles have evolved – AWS Certified Machine Learning Engineer – Associate (MLA-C02), AWS Certified Solutions Architect – Professional...
020
Elizabeth Fuentes L @elizabethfue12.bsky.social · 02/09/2026
AWS is credentialing the people who decide which AI projects ship and how to scale them. AWS Certified AI Business Strategist launches Sept 1. Tests business judgment, not tech implementation. #AI #AWS
ift.tt
Introducing AWS Certified AI Business Strategist: Built for the people who scale AI | Amazon Web Services
Every AI initiative has two kinds of work happening at once. There’s the building: the models, the infrastructure, the code. And there’s the deciding: which investments to fund, how to measure...
020
Elizabeth Fuentes L @elizabethfue12.bsky.social · 31/08/2026
Graviton5 R9g instances are GA: 25% faster compute, 5x larger L3 cache, DDR5-8800 (fastest memory in the cloud). Drop-in upgrade from R8g. #Graviton5 #AWS ift.tt/JAYnkUO
000
Elizabeth Fuentes L @elizabethfue12.bsky.social · 31/08/2026
AWS acquired DuckLabs. DuckDB stays open source (MIT license, independent foundation) and will integrate with S3, Redshift, and Athena. It's fast at the everyday queries most teams actually run and works well with AI agents. #DuckDB #OpenSource ift.tt/TL5e78R
020
Elizabeth Fuentes L @elizabethfue12.bsky.social · 30/08/2026
I'm judging a hackathon. $40k in prizes, deadline Sep 14. Build AI agents that do the boring stuff so you don't have to. Three tracks: everyday life, work productivity, or helping groups. Beginner friendly. 5,786 people registered. agentsforhumans.devpost.com
031
Elizabeth Fuentes L @elizabethfue12.bsky.social · 29/08/2026
Your AI agent took 30 seconds and 5 tool calls. You don't know why. New video on OpenTelemetry for agents: traces show the decision path, metrics track tokens and timing. www.youtube.com/watch?v=CvoO...
000
Elizabeth Fuentes L @elizabethfue12.bsky.social · 28/08/2026
Vector search returns disconnected facts. Graph memory walks relationships. "Who do I know connected to flights to Spain?" Graph: Maya → Iberia → Madrid → Spain Demo with Strands Agents + Neo4j dev.to/aws/graph-me...
100
Elizabeth Fuentes L @elizabethfue12.bsky.social · 26/08/2026
AWS Skill Builder Lab Maker now lets you save AI-generated custom labs and share them across your team with a URL. No more rebuilding the same lab every time. Create once, reuse forever. #AWS #CloudLearning
ift.tt
Announcing Lab Maker Build, Save, and Share: Create custom AWS Labs and share them across your organization | Amazon Web Services
Today, we’re excited to announce Build, Save, and Share for Lab Maker on AWS Skill Builder. With this launch AWS Skill Builder Team Subscription customers can now save their AI-generated custom labs,...
000
Elizabeth Fuentes L @elizabethfue12.bsky.social · 26/08/2026
AWS Card Clash now has Contests (async leaderboard challenges) and Tournaments (live PvP brackets). Contests rank by best attempt per architecture, not total replays. Smart scoring. #AWS #CloudLearning
ift.tt
Introducing Contests and Tournaments: New ways to practice and compete in AWS Card Clash | Amazon Web Services
Cloud learning is more effective when learners can apply what they have learned, compare approaches with others, and choose an experience that fits the moment. Sometimes a team needs a flexible challenge...
000
Elizabeth Fuentes L @elizabethfue12.bsky.social · 25/08/2026
EC2 is 20 years old today. Started with one instance type in one Region. Now it's 1,200+ types across 39 Regions, plus custom silicon for AI workloads and actual Mac minis running in AWS data centers. #AWS #EC2 ift.tt/26Pu3Fg
000
Elizabeth Fuentes L @elizabethfue12.bsky.social · 24/08/2026
Verified students: 12 months AWS Skill Builder premium free, then earn badges for activity. 21 badges = $100 cert exam voucher. Also new: Las Vegas Local Zone, batch instance termination (up to 100 at once), CloudShell visual editor. #AWS #CloudComputing ift.tt/N1M9zgR
000
Elizabeth Fuentes L @elizabethfue12.bsky.social · 22/08/2026
DynamoDB now has vector search built in. Single-digit ms, same accuracy as S3 Vectors, vectors in the same table. buff.ly/LBxCRpO
dev.to
Amazon DynamoDB Vector Search. No Separate Vector Store
Amazon DynamoDB Vector Search GA: add a vector index to an existing table. Same accuracy as S3 Vectors, single-digit ms latency, data and embeddings collocated.
110
Elizabeth Fuentes L @elizabethfue12.bsky.social · 21/08/2026
AWS Glue 6.0: 30% cheaper + full Iceberg v3 support. VARIANT data type with shredding means you can query JSON without flattening schemas. Arrow-native Python UDFs eliminate JVM serialization overhead. #AWSGlue #ApacheIceberg ift.tt/k6pC7A9
000
Elizabeth Fuentes L @elizabethfue12.bsky.social · 18/08/2026
AWS is opening permanent Builder Lofts in Berlin, Hyderabad, and São Paulo. Free workshops, hackathons, and space for local dev groups to host meetups. SF's loft hit 22,500 devs since July. #AWS #DevCommunity ift.tt/CS7jrVn
000
Elizabeth Fuentes L @elizabethfue12.bsky.social · 17/08/2026
EC2 now checks if your app is actually working, not just if the instance is up. Detects stopped web servers, dead Docker daemons, broken network configs. Finally. #AWS #EC2 ift.tt/Hp41GEA
010
Elizabeth Fuentes L @elizabethfue12.bsky.social · 15/08/2026
DynamoDB now has native vector search. Single-digit millisecond latency at 99%+ recall, no separate vector database, no sync pipeline. Store embeddings as floats, create a vector index, run SearchVectors. #DynamoDB #VectorSearch
ift.tt
Amazon DynamoDB now supports real-time vector search at any scale | Amazon Web Services
DynamoDB now supports native vector search with single-digit millisecond latency at 99%+ recall. It is designed for any scale, even trillions of vectors and requires zero infrastructure management.
000
Elizabeth Fuentes L @elizabethfue12.bsky.social · 13/08/2026
Your AI agent doesn't need a vector database. It needs semantic search. Built with Strands Agents: keywords miss, embeddings find. dev.to/aws/do-ai-ag...
010
Elizabeth Fuentes L @elizabethfue12.bsky.social · 12/08/2026
AWS Cloud Quest now starts every assignment with an AI-powered customer conversation. You ask questions, diagnose the problem, then build the fix in a live AWS environment. This is what learning by doing actually looks like. 🛠️ #AWS #CloudLearning
buff.ly
Introducing the new AWS Cloud Quest: AI-powered practice, hands-on building, and a path to official AWS badges | Amazon Web Services
Today we're announcing a major refresh to AWS Cloud Quest, the 3D, role-based learning game on AWS Skill Builder.
000
Elizabeth Fuentes L @elizabethfue12.bsky.social · 10/08/2026
OpenAI models on Bedrock now search the web in real time. pull live info while staying in your AWS environment with zero data egress. #Bedrock #GenAI ift.tt/R21ZGyg
000
Elizabeth Fuentes L @elizabethfue12.bsky.social · 07/08/2026
Runtime instances in Amazon Bedrock AgentCore: agents that run for up to 14 days, GPU support, multi-agent collaboration on the same host. Managed infrastructure so you don't build it yourself. Bring any framework, any model. 🤖 #GenerativeAI #AWSBedrock ift.tt/RZV3GIq
000
Elizabeth Fuentes L @elizabethfue12.bsky.social · 05/08/2026
Your AI agent doesn't remember users. The transcript is re-sent every turn and dies on restart. Key-value memory fixes it: store named facts, persist them, restore on a fresh instance. No vector DB needed. dev.to/aws/stop-you...
000
Elizabeth Fuentes L @elizabethfue12.bsky.social · 03/08/2026
GPT-5.6 Luna on Bedrock: $0.20 per million input tokens after an 80% price cut. Frontier models just got a lot cheaper to run at scale. #GenerativeAI #AWS ift.tt/QEIVguy
010
Elizabeth Fuentes L @elizabethfue12.bsky.social · 31/07/2026
AWS launched 3 free data engineering microcredentials: streaming, lakehouse, viz. No multiple choice. Live environment, real problems, timer. You fix it or you don't. #DataEngineering #AWS
ift.tt
Prove your data analytics skills: Three new microcredentials now available | Amazon Web Services
Today, we're launching three new microcredentials focused on Data Analytics: AWS Data Streaming Demonstrated, AWS Data Lakehouse Demonstrated, and AWS Data Visualization Demonstrated. Each one validates your...
011
Elizabeth Fuentes L @elizabethfue12.bsky.social · 29/07/2026
AWS Skill Builder Trivia added async mode: 30-day windows, same speed scoring (1k points max), no live host required. Plus polls for opinion questions. Works for distributed teams that can't sync schedules. #AWS #CloudLearning
ift.tt
Introducing Self-Paced Trivia and Polls: new ways to compete and connect in AWS Skill Builder Trivia  | Amazon Web Services
Today, we’re excited to announce Self-Paced Trivia and Polls, two new capabilities that give teams more flexibility in how they learn and engage. Self-Paced Trivia lets participants compete...
000
Elizabeth Fuentes L @elizabethfue12.bsky.social · 25/07/2026
New Tech & AI Weekly is out. The bottleneck in AI agents isn't the model anymore, it's everything around it: context, GPU capacity, memory. Plus a judge put a $1.5B price tag on training data. blog.elifuentes.tech/blog/newslet...
010