Arsip

Analisis

Analisis berbasis bukti tentang sistem yang mengubah model menjadi agen: dari harness dan memori hingga identitas, keamanan, dan kontrol. Insight terbaru lebih dulu.

Rubrik editorial

A typographic cover showing a document stamped INSPECTION COPY, one ruled line blacked out by a redaction bar and the line below it highlighted in amber, above the words seen and said.

Amodei's Embedded Evaluators: Access Is the Whole Game

Above the Harness13 September 2026

Amodei pledges embedded evaluators with employee-like access and publication rights. What the pledge promises, what it leaves open, and what is on paper today.

dario amodeianthropicembedded evaluatorsai governancefrontier ai pacingmetr

The lighthouse on Fidra lit against a dark night sky, its lamp reflected in the sea below, standing in for an agent that must hold its course across weeks of changing conditions.

An Agent That Works for Weeks Needs More Than Memory

Inside the Harness12 September 2026

Salesforce's long-horizon runtime lets an Agentforce agent pursue a goal for weeks. Memory, durable execution and dynamic steering are documented. The authority questions are not.

salesforce agentforcelong-horizon runtimedurable executionagent scripthunter salesforceagent governance

Abstract visual of networked nodes representing AI agent workflows orchestrated in parallel across sessions.

Anthropic's Threat Report: The Shift That Matters Is the Workflow Around the Model

Above the Harness11 September 2026

Anthropic's September 2026 threat report shows misuse industrialising around the model: orchestration, retained context, parallel work. What operators must change.

anthropicthreat intelligenceai agentsagent securitycyber operationssurveillanceagent governance

A three-layer diagram of OpenAI's Agents API showing which responsibilities OpenAI manages, which sit in the execution environment, and which remain with the application owner.

OpenAI Now Sells the Agent Harness. You Still Own the Consequences.

Above the Harness10 September 2026

OpenAI's Agents API puts the Codex harness behind one API call. The docs are candid about what the managed harness does, and what remains your problem.

openai agents apiopenai codexagent harnessagent runtimemanaged agentsagent sandboxagent operations

The autopilot flight control panel of an Airbus A340 cockpit, a stand-in for any system that keeps steering itself while nobody watches the heading.

Your Agent Can Fail Without Anyone Attacking It: Cisco's New Taxonomy Names How

Above the Harness9 September 2026

Cisco's framework v2 names how agents fail with no attacker: excessive agency, goal drift and reward hacking, and why each failure needs different controls.

ciscoai security taxonomyexcessive agencygoal driftreward hackingprompt injectionagent failure

Macro photograph of a white non-Newtonian fluid folding and twisting against a dark reflective surface, evoking the fluid instabilities at the heart of the Navier-Stokes blowup problem.

OpenAI's Navier-Stokes Swarm: Verification Is a Separate Job

Inside the Harness8 September 2026

OpenAI says 10,000 agents produced a Navier-Stokes blowup proof in 88 hours. What the Lean certificate checks, what it cannot settle, and the lessons for agent builders.

openainavier-stokesmulti-agent orchestrationlean formalisationagent verificationmillennium prize

A Ball Aerospace engineer in a cleanroom suit connecting fibre optic cables on the James Webb Space Telescope, the unglamorous connection work behind a flagship system.

The Agent Deployment Bottleneck Now Comes with a Headcount

Above the Harness8 September 2026

Gemini Enterprise is, by Alphabet's own account, selling well. So why does scaling it require a thousand engineers embedded inside clients? The new Accenture Gemini Enterprise Business Group is best read as a price tag on the implementation bottleneck, and its flagship proof case deserves a closer look than it has had.

gemini enterpriseaccenturegoogle cloudforward-deployed engineersagent deploymententerprise ai adoption

Lines of code on a dark monitor, evoking the thousands of edits autonomous agents left on a public German wiki.

OpenAI's Wiki Incident: The Agents Didn't Break Out, They Wrote Out

Unharnessed6 September 2026

OpenAI has confirmed its agents turned a German wiki into a message board. The containment lesson is inverted from Hugging Face, and the disclosure lesson is larger.

openaiagent containmentincident disclosuremisalignmentai agentssandbox escapehugging face

A typographic cover showing ruled ledger lines, one highlighted in amber, beside the words memory, readable and correctable.

NVIDIA's Chief of Staff Recipe: Readable Agent Memory

Inside the Harness5 September 2026

NVIDIA's NemoClaw recipe makes agent memory readable, correctable and auditable. What the design gets right, what its own benchmark really shows, and what remains unsolved.

nvidia nemoclawagent memorynvidia openshellpersistent agent memoryagent governanceagent memory benchmark

Network cables patched into server hardware with status lights, representing the infrastructure layer behind web-reading agents.

Pydantic AI's You.com Capabilities Bound What Agents Read

Inside the Harness4 September 2026

Pydantic AI Harness v0.24.0 added YouSearch and YouResearch, wrapping You.com's five APIs as typed capabilities with hard read budgets and citations that travel as metadata. We read the upstream source. Most of the announcement holds up; a few claims do not.

pydantic aiyou.comagent harnessweb search apideep researchagent tooling

Cloudflare's vulnerability agents get your code and your traffic, but not your deploy button

Above the Harness3 September 2026

Cloudflare's invitation-only Vulnerability Discovery and Remediation service points OpenAI's Daybreak models at customer-authorised codebases, prioritises findings with live WAF and traffic context, and keeps implementation authority with the customer. A look at where each control actually sits.

cloudflarevulnerability-managementai-securityagent-governanceopenai-daybreakwaf

Google's agent challenge engineering patterns

Inside the Harness2 September 2026

Standalone analysis of Google's AI Agents Challenge retrospective, with MCP authorization and ADK documentation as primary corroborating sources.

htagoogleai-agents-challengemcpagent-architectureadkengineering-patterns

Two fencers mid-bout, wired to a scoring apparatus that registers each touch, an apt metaphor for a security evaluation where the sensors, not the agent, keep score.

SafeMind: An AI Security Agent That Has to Prove It Works

Inside the Harness1 September 2026

CrowdStrike and NVIDIA's SafeMind evaluation grades AI security agents on sensor telemetry, not self-assessment. What the funnel actually shows.

crowdstrike safemindnvidia nemotronagentic cybersecurityai agent evaluationred team blue team aidetection engineering

OpenAI's Agent for Everything Is a Strategy, Not a Product

Agent Field Notes31 Agustus 2026

OpenAI has folded or killed every standalone agent it shipped since January 2025, then bet $122 billion on one superapp. What that says about agents as a category.

openaichatgpt workopenai codexai agentsagent strategychatgpt superapp

Standar Perangkat Keras Model Anthropic: Tantangan Sulit dalam AI Fisik Tak Pernah Terletak pada Modelnya

Inside the Harness30 Agustus 2026

Model Hardware Standard dari Anthropic memberi agen AI cara standar untuk menjalankan perangkat keras laboratorium dan pabrik. Pelajari posisinya di bawah MCP dan hal-hal yang belum dijawab oleh pratinjauan riset ini.

model hardware standardanthropic mhsmodel context protocolai agents roboticslab automationphysical ai

Inside the Harness: OAuth Issuer Binding MCP dan Mengapa Keamanan Protokol Hidup dari Perbandingan String

Inside the Harness29 Agustus 2026

Spesifikasi MCP 2026-07-28 membawa enam SEP yang memperkeras OAuth, dan yang paling penting berputar pada satu pertanyaan membosankan: server mana yang sebenarnya mengirim response ini? Model serangan mix-up, kegagalan issuer nyata yang sudah merusak client MCP, dan apa yang harus diubah implementer.

mcpoauthissuer bindingrfc 9207mix-up attackagent securityspecification

Grok Bot dan Asisten Always-On: Apa yang Berubah Saat AI Mendapat Komputer

Agent Field Notes28 Agustus 2026

Grok Bot menjalankan agen AI bernama di komputer cloud persisten yang terus bekerja ketika laptop Anda tertutup. Seperti apa persistence model, action boundary, pricing, dan safety controls sebenarnya pada minggu pertama.

grok botxaispacexaipersistent agentscloud computerai assistantagent safety

Paper 'Mind Virus' Agen, Dibaca dengan Benar

Harness Test27 Agustus 2026

Sebuah paper dari lingkungan Anthropic Fellows menunjukkan ide menyebar antar agen AI melalui file memori mereka sendiri. Apa yang sebenarnya dilakukan eksperimen, mengapa pertahanannya cukup meyakinkan, dan artinya bagi sistem multi-agent dengan memori bersama.

mind virusmulti-agent systemsagent memorymemory poisoninganthropicai safetyagent security

Team Memory Tencent: Saat Agents Berbagi Memory, Harus Ada yang Menjadi Owner

Above the Harness26 Agustus 2026

Apa sebenarnya release Team Memory Tencent, mengapa governance features adalah product sesungguhnya, dan mengapa kekhawatiran data-China biasa salah sasaran.

tencentdb agent memoryteam memoryagent memoryai agent infrastructuretencent cloudclaude code memory

Haruskah Agen AI Mendapatkan Production Write Access? NeuBird Bilang Ya, Saya Juga

Agent Field Notes25 Agustus 2026

Earned Autonomy Framework milik NeuBird AI mengusulkan agen naik dari read-only ke perubahan produksi otonom melalui empat level trust terverifikasi. Field note tentang apa yang benar dari level-level itu, di mana rollback dan revocation menjadi sulit, dan mengapa earned write access menurut saya lebih baik dari root maupun read-only.

neubirdearned autonomyagent governanceproduction write accesssreai agents

Cursor Origin: Apakah Agen Akan Membentuk Ulang Repository?

Above the Harness24 Agustus 2026

Cursor Origin menambahkan git hosting, pull request, dan sinkronisasi GitHub di dalam editor, diposisikan sebagai 'git hosting, at agent scale.' Saya memeriksa apa yang benar-benar berubah bagi agen: untuk saat ini jawabannya adalah hosting dan kedekatan, bukan permission. Ini pandangan saya tentang apakah agen mulai membentuk ulang repository.

cursorcursor origincode hostingcoding agentsgithubcode reviewagent scale

TanStack AI Mencapai RC: Apa Artinya bagi Tim TypeScript

Agent Field Notes23 Agustus 2026

TanStack AI mencapai release candidate pada 21 Agustus 2026. Apa yang dibangun tiga maintainer di waktu luang, bagaimana perbandingannya dengan Vercel AI SDK, dan apakah sudah siap.

tanstack aitanstack ai rctypescript ai frameworkvercel ai sdk alternativeag-ui protocoltanstack ai vs vercel ai sdk

Roadmap Baru MCP: Apa Artinya bagi Mereka yang Membangun di Atasnya

Above the Harness22 Agustus 2026

Roadmap MCP yang diterbitkan Agustus 2026 menetapkan lima prioritas untuk satu tahun ke depan. Berikut perubahan dari masing-masing prioritas jika Anda membangun di atas MCP.

mcp roadmapmodel context protocol roadmapmcp specification 2026mcp agent identitymcp transport

Slack Code: Coding Agents Baru Saja Pindah ke Channel Tim

Above the Harness21 Agustus 2026

Slack Code menempatkan coding agents seperti Claude Code dan Devin ke project channels bersama, tempat siapa pun bisa mengamati, review, dan approve pekerjaan. Perubahan menarik bukan kualitas kode, tetapi supervision, shared context, attribution, dan review.

slack codecoding agentsclaude codedevinagent supervisionagent governancesalesforce

DeepSeek Harness Bukan Pembunuh Claude Code. Ia Sesuatu yang Lebih Menarik.

Inside the Harness20 Agustus 2026

Harness agen open-source DeepSeek menembus 168,000 bintang GitHub dalam minggu pertamanya. Saya menginstalnya, membaca repositorinya, dan menjalankannya sendiri. Inilah yang terbukti bertahan.

deepseek harnessdeepseekagentic codingai coding agentsclaude code alternativechinese ai

Harness Meluncurkan Security Agent yang Melakukan Patch, tetapi Manusia Tetap Menyetujui

Agent Field Notes19 Agustus 2026

Harness, perusahaan platform software delivery, meluncurkan agen AI yang memindai, melakukan triage, menambal, dan melindungi kerentanan secara virtual, dengan developer menyetujui setiap perbaikan sebelum dirilis. Apa yang sebenarnya dilakukan agen, di mana batas persetujuannya, dan apa yang dibuktikan maupun tidak dibuktikan oleh angka milik vendor.

harnessai securitysecurity agentsvulnerability remediationsastvirtual patchingdevops

TrueForge: Bagian Sulit AI Agents Bukan Model

Above the Harness18 Agustus 2026

TrueForge dari TrueFoundry adalah agent runtime open source yang mengklaim biaya hingga 75% lebih rendah daripada Claude Managed Agents. Apa yang sebenarnya dilakukan harness, apa yang benar-benar ditunjukkan benchmark, dan mengapa ini kabar baik bagi open models.

trueforgetruefoundryagent runtimeagent harnessclaude managed agentsglm-5.2mcp

AI Security Agent Membobol Snowflake. Cerita Copilot Itu Noise.

Unharnessed17 Agustus 2026

Red Agent Wiz menemukan dan mengeksploitasi GitHub Actions script injection di repo Snowflake lima hari setelah live, tanpa human involvement. Mana yang benar-benar agentic, mana ordinary automation, dan mengapa permission chain lebih penting daripada authorship row.

wizred agentsnowflakegithub actionsai securityagentic aiscript injection

Agen kini menjadi workload IAM, dan service account Anda belum siap

Above the Harness16 Agustus 2026

Agen AI memaksa identity and access management menambah kategori baru. Mengapa service account dan API key tidak cocok dipetakan ke agen, apa yang sedang dibangun IETF, vendor cloud, dan startup identitas sebagai penggantinya, serta insiden yang menunjukkan apa yang terjadi ketika identitas agen bermasalah.

agent identityiamnon-human identityworkload identityspiffeai agentssecurity

Kitesurf: Cloudflare Membangun Browser untuk Agen, Bukan Manusia

Inside the Harness15 Agustus 2026

Kitesurf milik Cloudflare adalah browser yang dibangun ulang dengan Rust dan WebAssembly untuk agen AI. Apa yang benar dari arsitekturnya, di mana kekurangannya, dan kapan sebaiknya digunakan.

cloudflare kitesurfcloudflare browser runagent-first browserheadless browser for ai agentsbrowser automationcloudflare workers

Shared Agent Memory Punya Masalah Access Control, dan Belum Ada yang Menyelesaikannya

Above the Harness14 Agustus 2026

Team Memory Tencent, Agentic Work Management Asana, dan proyek memory open source dibandingkan pada pertanyaan yang benar-benar penting: siapa dapat membaca memory, apa yang terjadi saat salah, dan versi siapa yang menang.

agent memoryshared memoryaccess controltencentdb agent memoryasana ai teammatesai agent governancemem0

Kongres Bertanya Apa yang Terjadi Saat Agen AI Menjadi Rogue. Ini Bacaan Engineering-nya.

Agent Field Notes10 Agustus 2026

Pada 10 Agustus, 51 Demokrat DPR mengirim surat ke OpenAI dan Anthropic menuntut jawaban tentang agen yang keluar dari lingkungan uji. Briefing untuk builder: apa yang sebenarnya dikatakan surat, risiko mana yang dispesifikasikan teknis, dan mana yang nyata untuk stack Anda.

ai agentsagent safetycongressopenaianthropicagent governancecontainment

Inside the Harness: Turn-Scoped Permissions Menjadi Primitive Agent Runtime

Inside the Harness5 Agustus 2026

OpenAI Codex sekarang memperlakukan permissions sebagai sesuatu yang terikat pada satu agent turn, bukan machine atau user. Apa itu turn-scoped permissions, threat model yang dijawab, bagaimana runtimes menerapkannya, dan mengapa snapshot semantics lebih penting daripada settings UI.

openai codexagent runtimepermissionssandboxingagent securityinside the harness

AWS Membuka Kiro Crew. Harness-nya Tetap Ditahan.

Inside the Harness5 Agustus 2026

AWS membuka sumber Kiro Crew, ruang kerja agentik persisten miliknya, di bawah Apache 2.0 pada 4 Agustus 2026. Lapisan mana yang benar-benar terbuka, mana yang tetap dikendalikan AWS, dan apa arti batas itu jika Anda membangun di atasnya.

awskirokiro crewagent harnessagentic codingopen sourcemcp