{"id":34790,"date":"2026-08-29T14:22:56","date_gmt":"2026-08-29T12:22:56","guid":{"rendered":"https:\/\/www.angulararchitects.io\/blog\/ae-summer-2026-update-for-angular\/"},"modified":"2026-09-03T16:09:51","modified_gmt":"2026-09-03T14:09:51","slug":"ae-summer-2026-update-for-angular","status":"publish","type":"post","link":"https:\/\/www.angulararchitects.io\/en\/blog\/ae-summer-2026-update-for-angular\/","title":{"rendered":"Agentic Engineering: Summer ’26 Update"},"content":{"rendered":"
\n\t\t\t\n\t\n\t\n\t\t

\n\t\t\tThis is post 8 of 9 in the series “Agentic Engineering”<\/em>\t\t<\/p>\n\t\t\t<\/label>\n\n\t\t\t

\n\t\t\t
    \n\t\t\t\t\t\t\t\t\t
  1. Agentic Engineering: Which LLM Is Best for Angular Development?<\/a><\/li>\n\t\t\t\t\t\t\t\t\t
  2. Agentic: Which App\/Harness Is Best for Angular Development?<\/a><\/li>\n\t\t\t\t\t\t\t\t\t
  3. Agentic Engineering: What Does AI Coding Really Cost?<\/a><\/li>\n\t\t\t\t\t\t\t\t\t
  4. Agentic Engineering: What Do AI Coding Tools Do With Your Code?<\/a><\/li>\n\t\t\t\t\t\t\t\t\t
  5. Agentic Verdict: What’s the Best Solution for You?<\/a><\/li>\n\t\t\t\t\t\t\t\t\t
  6. Anthropic Just Released the First Next-Gen Model to the Public<\/a><\/li>\n\t\t\t\t\t\t\t\t\t
  7. Agentic Engineering: How to Set Up Your Harness for Angular Development<\/a><\/li>\n\t\t\t\t\t\t\t\t\t
  8. Agentic Engineering: Summer ’26 Update<\/span><\/li>\n\t\t\t\t\t\t\t\t\t
  9. Agentic Engineering: Skills for Angular Development<\/a><\/li>\n\t\t\t\t\t\t\t<\/ol>\n\t\t<\/div>\n\t<\/div>\n

    My agentic engineering series<\/a> started in late May 2026, and I promised myself not to turn this blog into a weekly model-news ticker. But this space moves so fast that after one summer, some of my recommendations have quietly changed \u2013 and I'd rather tell you in one short post than let the old ones age in silence. So this is a quick end-of-summer status update \ud83d\udccb \u2013 the first in what I plan to make a quarterly series \u2013 across the three layers of the stack: the models<\/strong>, the harness<\/strong>, and the apps<\/strong> \u2013 plus what all of that costs<\/strong>. If you're new here, the apps and harnesses post<\/a> and the harness setup post<\/a> explain how I slice those layers. This is a subjective field report from my daily work.<\/p>\n

    TL;DR: What Changed This Summer<\/h2>\n

    If you only have thirty seconds: my preferred models are now Fable 5<\/strong>, GPT-5.6 Sol<\/strong>, and \u2013 as the fast, cheap, and open-weight pick \u2013 GLM-5.3 Flash<\/strong>. On the harness layer, nothing important changed \u2013 the setup recommendations<\/a> hold up. On the app layer, however, there is real news: the open-source T3 Code<\/strong> is now my preferred app, with the Codex app<\/strong> as my alternative for computer use. Every layer but the apps gets the latest DeepSWE<\/strong> numbers \u2013 official for models and costs, via Artificial Analysis for harnesses \u2013 and each one gets a small podium<\/strong> with my three personal favorites. And the one thing to take home is my current workflow: Fable 5 as the orchestrating agent, GPT-5.6 Sol as the sub-agent \u2013 the best way I know to stretch the precious Fable budget. Plus my two cents on loop engineering<\/strong>, the buzzword of the summer.<\/p>\n

    Models: My New Trio<\/h2>\n

    In the LLM post<\/a> from late May, my ranking was Opus 4.7 for Angular<\/em>, GPT-5.5 as the all-rounder, and Composer 2.5 as the newcomer. One summer later, all three have been pushed aside: Opus 4.7 by its maker's next generation, GPT-5.5 by its own successor, and Composer 2.5 by a model I hadn't even heard of in May \u2013 not because they got worse, but because the next round arrived. Interestingly, Cursor \u2013 by now part of SpaceX, since the SpaceXAI deal from the apps post closed on August 14 \u2013 didn't produce a podium model at all. In the apps post<\/a> I expected them to enter the frontier stage, and so far that call was wrong.<\/p>\n

    1: Fable 5 by Anthropic<\/h3>\n

    No surprise if you read my Fable 5 post<\/a>: this model<\/a> is a new generation<\/strong>, and it is still the best model I have ever pointed at an Angular<\/em> codebase \u2013 or at anything else, by the way. It is very intelligent and \u2013 I keep coming back to this word \u2013 expansive<\/strong>: give it a vague goal and it explores the problem space, tests its own ideas, and comes back with validated results instead of a guess. And the best part: it usually just understands what I meant. After the June shutdown drama, access came back on July 2 and has stayed ever since \u2013 and Fable 5 has settled in as my go-to for everything hard: architecture, design, reviews, and long-horizon refactorings. The catches from that post \u2013 price, speed, usage limits \u2013 are all still real, which is exactly why the sub-agent workflow at the bottom of this post exists.<\/p>\n

    2: GPT-5.6 Sol by OpenAI<\/h3>\n

    In the Fable post, I predicted OpenAI would answer with GPT-6<\/strong>. They answered with GPT-5.6<\/a> instead \u2013 and honestly, the name is the least interesting part. The big variant, GPT-5.6 Sol<\/strong>, is very intelligent, clearly a step up from GPT-5.5, and it comes at a fair price<\/strong>: the included usage in the OpenAI subscriptions remains generous, so I can let it work all day without watching a meter. And on August 21, OpenAI cut the Sol price<\/a> again \u2013 20% on input tokens, a third on output tokens, valid through November 21 \u2013 the third price cut in the GPT-5.6 family within a month; the subscriptions stay as they are, but Codex credits and API calls get cheaper too. It has taken over the role GPT-5.5 had in my setup \u2013 the reliable senior engineer for focused implementation work \u2013 and, as you'll see below, it is my sub-agent of choice. Both Fable and Sol write good, mergeable code, but Sol needs a bit more steering and is definitely not as capable as Fable.<\/p>\n

    3: GLM-5.3 Flash by Zhipu (Z.ai)<\/h3>\n

    The third step of my podium was empty for most of the summer. In May, this was the newcomer slot, held by Composer 2.5 \u2013 which, next to what came after it, simply isn't capable enough anymore. Then I gave Grok 4.5<\/strong><\/a> a real shot at it \u2013 cheap, and for mechanical refactorings and test generation it held up fine \u2013 but when it mattered, I kept reaching for the two models above.<\/p>\n

    What finally filled the slot came from a lab I wasn't even tracking in May: GLM-5.3 Flash<\/strong><\/a>, the small sibling of Zhipu's GLM-5.3<\/a>. It is not frontier intelligence \u2013 63% on DeepSWE, ten points behind Sol \u2013 but it is absurdly fast<\/strong>, it gets there for about \u20ac0.25 per task, and in my Angular<\/em> work it is simply more capable than the obvious alternative, GPT-5.6 Luna, which I already pay for through the OpenAI subscription.<\/p>\n

    The leaderboard doesn't back that up outright \u2013 Luna at max scores 67% \u2013 but at the effort I actually run Luna, high, it drops to 44%, while the Flash gets its 63% flat out at max, and at that price, effort levels stop mattering. That makes it the cheapest model whose output I'd actually merge. For the mechanical share of my work \u2013 the well-specified legwork you don't need to hand to Fable anyway \u2013 that is often enough, and next to two expensive models, a dirt-cheap third one is exactly what my setup was missing. More on that in the costs section, where it climbs to the top of the podium.<\/p>\n

    The shortlist behind it is longer than ever, and I'll keep re-evaluating it every quarter:<\/p>\n