BLS 2025-35: data science roles +34.6%, research scientists +21.8%.
Admin support: -4.0%, 752,100 jobs lost.
BLS: AI gains may dampen demand for some roles.
Know which role software changes before pricing savings.
#JEBREX#AIResearch#Productivity#Automation#AIAdoption
Anthropic gave Agent Skills, its open standard for agent know-how, to the Agentic AI Foundation on Oct 22.
A skill is one folder, one SKILL.md file.
1M+ published, near 280M installs.
Write yours as files you own.
#JEBREX#AgenticAI#Anthropic#AIAgents#Automation
A support queue measures tickets closed today.
It says nothing about the account that never invited a second user, or the admin seat empty since setup.
Customer success starts reading usage, not inboxes.
#JEBREX#CustomerSuccess#SaaS#B2BSaaS#SaaSPricing
TRADREX shipped in phases: spec, safety filters, AI analysts, risk engine, paper trading.
Each phase gated the next: tests, then an audit, before moving on.
Findings got fixed where found, never deferred.
#JEBREX#TRADREX#PaperTrading#AlgorithmicTrading#BuildingSoftware
Most calls to a public service line are requests for status, not action.
Reciting "under review" to caller after caller isn't judgment. It's lookup.
Free that time for the case that needs a real decision.
#JEBREX#PublicSector#AIAdoption#DigitalTransformation#Automation
Google's ADK pairs an LlmAgent that picks its own path with workflow agents that just run a fixed sequence, loop, or parallel template.
Same toolkit. Autonomy becomes a per-step choice, not a whole-product bet.
#JEBREX#AgenticAI#AIAgents#WorkflowAutomation#Automation
A rival can clone every feature in your product by next quarter.
They can't clone the workflows wired into your system, or the history that lives nowhere else.
That's the moat. The feature list never was.
#JEBREX#CustomerRetention#ProductStrategy#VerticalSaaS#UnitEconomics
TRADREX's risk engine never takes the AI's word for a trade.
It recomputes every check itself, from fresh data - not what an earlier stage decided.
Missing data gets no benefit of the doubt. A gap is a no, never a guess.
#JEBREX#RiskManagement#DataQuality#PaperTrading#AI
A beta header just raised Claude's batch output cap to 300,000 tokens, standard pricing, as of Sept 2026, across Opus and Sonnet.
That fits work nobody's waiting on.
Two paths: fast for watching, patient for the rest.
#JEBREX#ContextWindow#AIModels#Anthropic#LLM
NVIDIA ACABA DE ELIMINAR EL MAYOR PUNTO DE FRICCIÓN EN VOICE AI
han liberado PersonaPlex-7B, un modelo de conversación dúplex completo que puede escuchar y hablar al mismo tiempo
→ no espera a que termines de hablar
→ usa una arquitectura de doble flujo para procesar y responder en tiempo real
→ código disponible en GitHub
y es 100% open source y gratis
os dejo el repo abajo
Claude's memory tool stores nothing itself. It requests file operations in a /memories folder, and your app executes them against storage you control.
Retention, encryption, deletion — your calls, not the model's.
#JEBREX#AIAgents#Anthropic#AIGovernance#SoftwareDevelopment
A person clicks a button. An agent acts for them. A cron job fires on its own.
Log all three under one shared API key, and you can't tell which one made the change.
Give each actor its own identity in the log.
#JEBREX#AIAgents#AIGovernance#APIDesign#SystemIntegration
Anthropic's Sept 2026 threat report: no malicious activity on Claude Fable or Mythos. Every disrupted op ran on Haiku, Sonnet, Opus.
Clean isn't safer. It just means nothing's surfaced yet.
Monitor every tier you expose.
#JEBREX#AISafety#Anthropic#AIInfrastructure#AIModels
A generated test suite that passes 100% proves nothing.
Flip one condition and rerun it. If nothing goes red, you don't have tests - you have confirmation bias with syntax highlighting.
#JEBREX#QualityAssurance#AICoding#SoftwareDevelopment#DevOps
Reading a page wrong is cheap. Clicking submit on what you misread is not.
Gate send, pay, and delete behind one confirmation step, even when everything before them runs on its own.
The irreversible step earns the pause.
#JEBREX#ComputerUse#AIAgents#AgenticAI#AIGovernance