IPv6 Press · Tech & Network Watch

GPT-6.1 Terra, not a broken price, but this one finally made up the middle.

$1 input / $5 output, more than Sol cut in half. But it's not the cheapest (and ten times cheaper Luna down) or the lower price — it's the replacement of an old model that has gone down. A four-storey price list, a credible ranking of six run points, and a security statement that nobody read out in that line.

itadol5j 2026-10-27

Microsoft puts AI Agent on the internal threat list: It's not the model. It's the authority.

Microsoft's framework on October 19 changed a default assumption: Agent is not one of us. There is only one sentence at the core — separating the supply of intelligence from the authority over it. In the same week, Agent across the street got the privilege in three hours.

itadol5j 2026-10-27

I went to check the CARLYOS report: four Claudes had a meeting, and refused to change the local code.

Straiker has access to the criminal gang's own work bank, 4,493 supporting documents, 200+ and zero-day leak reports. But the title "400 families broken down" and the original -- the original is 400+ targets.

itadol5j 2026-10-23

php2shell-audit field: Windows uses exe, Linux, py, 10 minutes to figure out PHP.

Read-only, no attack traffic, zero reliance on a single document. This is a self-censorship, a long-range batch, how to read the three conclusions, the four pits that have to be stepped on, plus an upgrade that must restart the worker.

itadol5j 2026-10-22

Java has a 9.1 split loophole, CVE-2026-91403, but I looked around the Oracle bulletin and found two big holes.

It's not in the CPU this season, and there's no RCE for that 9.1 cent. The three layers of Oracle ' s bulletin system, CVSS vector, supply chain are to be detached, so the problem is not in itself.

itadol5j 2026-10-22

PHP2Shell blew up: CVSS 9.2, the day before the opening, and the two holes were in code for 10 years.

CVE-2026-109618 and CVE-2026-109619, PHP 5.4 to 8.5 full line affected, without authorization for RCE-plus memory leaks. Patches, four interim mitigations, time lines, and the bug that was put on hold for two years.

itadol5j 2026-10-22

76.2 Ten thousand micro-letters are exported to AI: The hardest part is not AI

76.2 10,000 messages, 367.5 million words, 10 years of record. Export, cleaning, segmenting, quantifying, mixing, full process of five-step pipe, plus five pyres. And there's the fact that nobody wants to say it: it's not like it remembers a 1M context.

itadol5j 2026-10-20

Opencode-models-discovery 2.0 Release: Configure Shape Full, OpenCode v2 Formally Correct

Nine metadata formats, original/content vouchers, host hosting caches. The configuration field check and the four check lists moved from v1 to v2 are here.

itadol5j 2026-10-20

I spent half a month, Gemini 4 Argon.

First, Gemini 4 Argon is officially open today morning, October 13. I spent half a month cruising around for it, stepping on two walls in the middle, writing down below — by the way, where it was won, where it was lost, and showing it to you with an official run schedule. First wall: ...

itadol5j 2026-10-14

After the fire, the moat of Age Age software doesn't exist?

Let's start with a judgment that may not be very popular: "The code is complicated, so we can't copy it." The moat did collapse. But the "software without moats" is an over-reaction — one of the ones that collapsed and the one that was already the most vulnerable. A few things that have happened in the past week, just to crack this...

itadol5j 2026-10-13

SAST vs DAST vs IAST: Security test for complementarities of the three brothers, not substitution

The three most common abbreviations in the application security tests are SAST, DAST, IAST. Many people's first reaction was to “select one of the strongest”, but it was a hypocritical proposition — the three brothers were not competing, but viewed the same issue from different perspectives. It depends not on who's more advanced, but on who you're in...

itadol5j 2026-10-12

Why build a personal blog now?

The majority of people who do content now focus on the public sign, the voice, the little red book, the B stand. These platforms are high-flow and high-impact and appear to be “more cost-effective”. So, naturally, the question is: 2026, why is there a personal blog? I don't know.

itadol5j 2026-10-11