For years, programming something new followed a predictable, exhausting rhythm: write some code, hit a wall, and then disappear into a forest of documentation and StackOverflow tabs to find the trick to get it working. In 2025, that era ended for me.<\/p>\n\n\n\n
Today, AI code assistants take so much drudgery out of development and debugging that the work has become mostly gratifying and rarely frustrating \u2014 the opposite of how programming was in the Before Times.<\/p>\n\n\n\n
Between my work and my side interests, I often feel like I live <\/em>in Visual Studio Code (VSCode \u2014 a popular open-source development environment). Early in 2025 I subscribed to GitHub Copilot, which integrates AI coding assistants into VSCode. At $10\/month it’s a phenomenal bargain that offers software developers an easy way to loop in the latest models from OpenAI, Google, and Anthropic. Now when I’m trying something new (like this little project I did for fun<\/a>) I can mostly stay in VSCode and work with an AI assistant that has mastered all of the documentation.<\/p>\n\n\n\n As mentioned elsewhere<\/a>, my most significant side project in 2025 was helping Ukrainians develop an open-source ballistic calculator (pyballistic<\/a>) in Python and Cython. I polished that off at the end of September.\u00a0 By that point I had begun to spend more time with Github Copilot, and the capabilities of its latest models gave me enough confidence and support to tackle what would have previously been an absurdly ambitious project for my day job: an Excel Real-Time Data (RTD) server for the Interactive Brokers API.\u00a0 (This RTD server feeds live market data, positions, and orders directly into Excel using native Excel formulas.) After two months of working seven days a week on that, I had a beautiful piece of software so solid (and validated every build by over 800 unit tests) that I had begun to use it in live trading operations.<\/p>\n\n\n\n Watching these models mature over the last year has been like watching a child grow up.<\/p>\n\n\n\n Tell a child, “Clean your room.” First they’ll spend more time arguing than it would take to just do it. When they finally declare the task “done,” you might find a few toys picked up but most of the mess still there. Emphasize that “clean your room” means everything <\/em>and you might find the floor clean but everything shoved under the bed. <\/p>\n\n\n\n Claude v3 was notorious for hacking shortcuts. Ask it to fix a failing test and it might just replace the test logic with a “return true;” statement. Claude v3.5 wouldn’t be so brazen, but it was still prone to hack the example rather than the task. GPT-4 and Gemini v2 would enthusiastically announce completion without checking their work. Like the child who picks up two toys and concludes that his room must be clean, even though the mess is visible from outside the door.<\/p>\n\n\n\n The teens came quickly: Claude v3.7 and its contemporaries would often spend more effort arguing that its failure was actually success than it would have taken to do the work correctly.<\/p>\n\n\n\n More recent models have become more likely to keep checking and working until they succeed. Performance of the latest models is still wildly variable: a model that astonishes me with its apparent skill one day may choke on something relatively simple the next. But they are getting more consistent. And they are definitely getting more intelligent.<\/p>\n\n\n\n What is intelligence?<\/strong>\u00a0 It becomes easy to see when you’re doing hard work with different models. One of the neat things about Copilot is that you can choose to watch the model at work. They all think “out loud,” meaning you can read their chain of thought to understand how and why they do things. When it’s not having an off day, Claude Opus is intelligent<\/em>.\u00a0 Given a problem:<\/p>\n\n\n\n These are some of the things that separate a junior developer from a more experienced one. They are also qualities that characterize more intelligent <\/em>people.<\/p>\n\n\n\n Let me show you.<\/strong> Have you ever wondered what it’s like debugging software? Well, debugging <\/em>is one thing the newer models can usually do as well as a good human programmer. In fact, they can do it better because they can run the process faster and interact with the code more directly. Below I pasted a transcript of Claude working to find and fix a tricky bug in my RTD server. This could just as well have been a transcript of my thoughts if I had to debug it. But whereas this would have been a draining hour+ distraction for me, the Claude instance cranked this out in minutes.<\/p>\n\n\n\n\n\n\n\n \u2026 Now I see the issue. Let me check how the order list is populated – specifically the Read file:OrderManager.cs, lines 165 to 225<\/em><\/p>\n I see – when a new order list topic is created, it seeds with Searched for text Now let me look at Read file:OrderManager.cs, lines 880 to 960<\/em><\/p>\n Now I understand the flow. Let me check the test more closely to understand the potential race condition:<\/p>\n Read OrderBehaviorTests.cs, lines 1795 to 1845<\/em><\/p>\n I see a potential issue! The test subscribes to the order list AFTER calling But wait – the issue is Searched for text I see. When Since So the test should work\u2026 Let me think about this more carefully.<\/p>\n Actually, looking at the test, the issue is:<\/p>\n The problem is that Wait – the error message says the string is EMPTY! That means Let me look at Searched for text I found it! Look at line 314-316:<\/p>\n When Then when But wait – the intermittent failure suggests a race condition or ordering issue. Let me re-read line 198:<\/p>\n So it seeds if EITHER:<\/p>\n The test sequence:<\/p>\n Read OrderManager.cs, lines 775 to 835<\/em><\/p>\n I see! So the sequence in the test:<\/p>\n Read OrderInfo.cs, lines 180 to 220<\/em><\/p>\n Read OrderInfo.cs, lines 215 to 245<\/em><\/p>\n Now I understand the race condition! Here\u2019s what happens:<\/p>\n Hmm, that doesn\u2019t change the status. But the test is checking if \u201c12345\u201d appears in the list after Let me check what happens in Read OrderManager.cs, lines 865 to 895<\/em><\/p>\n Let me think about this more carefully. The test:<\/p>\n Unless\u2026 there\u2019s something odd happening with the order processing. Let me look at the test constructor call again:<\/p>\n Wait – let me check if there\u2019s an issue with empty strings in the initial seeding. The test reads Actually, I think I found the issue! Let me re-read the GetOrAddOrderListTopic flow:<\/p>\n If But when would Actually, this is a state leakage issue between tests! If a previous test left Let me look for something else. Is there perhaps a Configuration issue? Let me check what happens with Read OrderManager.cs, lines 60 to 90<\/em><\/p>\n The test uses Let me check Searched for text Early Childhood Development<\/h3>\n\n\n\n
\n
\n\n\n\nTranscript of Claude Debugging in VSCode<\/h2>\n\n\n\n
GetOrAddOrderListTopic<\/code> and HandleOpenOrderEnd<\/code> methods:<\/p>\nBuildOrderListCsv()<\/code> only if m_ordersSnapshotCompleted<\/code> is true. Let me check what HandleOpenOrderEnd<\/code> does:<\/p>\nHandleOpenOrderEnd<\/code>, 7 results. Read file:OrderManager.cs, lines 410 to 490<\/em><\/p>\nPublishOrderListsIfAny<\/code> and BuildOrderListCsv<\/code>:<\/p>\nHandleOpenOrder<\/code>, but BEFORE calling HandleOpenOrderEnd<\/code>. This is a problematic sequence because:<\/p>\n\n
HandleOpenOrder<\/code> adds orders to m_orders<\/code><\/li>\nGetOrAddOrderListTopic<\/code> is called – it sees m_ordersSnapshotCompleted = false<\/code> and m_ordersRequested = false<\/code>, so it seeds with empty string and calls EnsureOrdersRequested()<\/code><\/li>\nHandleOpenOrderEnd<\/code> sets m_ordersSnapshotCompleted = true<\/code> and calls PublishOrderListsIfAny()<\/code><\/li>\n<\/ol>\nm_ordersRequested<\/code>. Let me check:<\/p>\nm_ordersRequested<\/code>, 6 results. Read OrderManager.cs, lines 193 to 223<\/em><\/p>\nGetOrAddOrderListTopic<\/code> is called:<\/p>\n\n
m_ordersSnapshotCompleted = false<\/code> (orders have been added but HandleOpenOrderEnd<\/code> hasn\u2019t been called)<\/li>\nm_ordersRequested = false<\/code> (we haven\u2019t called EnsureOrdersRequested yet)<\/li>\n<\/ul>\n!m_ordersRequested<\/code> is true, the condition if (m_ordersSnapshotCompleted || !m_ordersRequested)<\/code> is TRUE, so it calls BuildOrderListCsv()<\/code> immediately!<\/p>\n\n
HandleOpenOrder<\/code> is called 3 times for different orders<\/li>\nGetOrAddOrderListTopic<\/code> is called – it seeds with BuildOrderListCsv()<\/code> because !m_ordersRequested<\/code> is true<\/li>\nHandleOpenOrderEnd<\/code> is called – it calls PublishOrderListsIfAny()<\/code> which updates the list again<\/li>\n<\/ol>\nBuildOrderListCsv()<\/code> checks ShowFilledOrders<\/code> which uses m_showFilledOrdersOverride ?? ShowFilledOrdersConfig<\/code>. The test uses showFilledOrdersOverride: false<\/code>, so that should be fine.<\/p>\nBuildOrderListCsv<\/code> returned an empty string. This could happen if:<\/p>\n\n
m_orders<\/code> is empty (but we just added 3 orders)<\/li>\nHandleOpenOrder<\/code>:<\/p>\npublic List<int> HandleOpenOrder<\/code>, 2 results. Read OrderManager.cs, lines 260 to 330<\/em><\/p>\n\/\/ Mark list as changed only if this is a new order\nif (isNewOrder)\n{\n \/\/ Update order list topics immediately (don't wait for next polling cycle)\n updated.AddRange(PublishOrderListsIfAny());\n}\n<\/code><\/pre>\nHandleOpenOrder<\/code> adds a NEW order, it calls PublishOrderListsIfAny()<\/code>. But when the test calls HandleOpenOrder<\/code> for the first time, there are NO order list topics yet (m_orderListTopics<\/code> is empty), so PublishOrderListsIfAny()<\/code> does nothing.<\/p>\nGetOrAddOrderListTopic<\/code> is called, if !m_ordersRequested<\/code> is true, it seeds with BuildOrderListCsv()<\/code>. This should work\u2026<\/p>\nif (m_ordersSnapshotCompleted || !m_ordersRequested)\n{\n var csv = BuildOrderListCsv(descriptor);\n<\/code><\/pre>\n\n
m_ordersSnapshotCompleted<\/code> is true (we\u2019ve finished the initial snapshot), OR<\/li>\n!m_ordersRequested<\/code> is true (we haven\u2019t requested orders yet – implies we\u2019re using pre-existing data)<\/li>\n<\/ol>\n\n
m_ordersRequested = false<\/code>, m_ordersSnapshotCompleted = false<\/code><\/li>\nHandleOpenOrder<\/code> 3 times – orders are added to m_orders<\/code><\/li>\nGetOrAddOrderListTopic<\/code> – m_ordersSnapshotCompleted = false<\/code>, m_ordersRequested = false<\/code>\n\n
!m_ordersRequested<\/code> is TRUE \u2192 calls BuildOrderListCsv()<\/code> \u2192 should return \u201c12345\u201d (only open order)<\/li>\nEnsureOrdersRequested()<\/code> which\u2026 let me check what that does<\/li>\n<\/ul>\n<\/li>\n<\/ol>\nEnsureOrdersRequested()<\/code> sets m_ordersRequested = true<\/code> and m_currentPollCycleUtc = now<\/code>.<\/p>\n\n
m_ordersRequested = false<\/code>, m_ordersSnapshotCompleted = false<\/code>, m_currentPollCycleUtc = DateTime.MinValue<\/code><\/li>\nHandleOpenOrder<\/code> 3 times – orders added to m_orders<\/code>, but m_currentPollCycleUtc<\/code> is still DateTime.MinValue<\/code> so orders are NOT marked as \u201cseen in poll\u201d<\/li>\nGetOrAddOrderListTopic<\/code>:\n\n
!m_ordersRequested<\/code> is TRUE \u2192 calls BuildOrderListCsv()<\/code> \u2192 should return \u201c12345\u201d<\/li>\nEnsureOrdersRequested()<\/code> \u2192 sets m_ordersRequested = true<\/code>, m_currentPollCycleUtc = now<\/code><\/li>\n<\/ul>\n<\/li>\nHandleOpenOrderEnd<\/code>:\n\n
m_ordersSnapshotCompleted = true<\/code><\/li>\nif (m_currentPollCycleUtc > DateTime.MinValue)<\/code> is TRUE (just set by EnsureOrdersRequested!)<\/li>\nif (orderInfo.LastSeenInPollUtc < m_currentPollCycleUtc)<\/code>\n\n
LastSeenInPollUtc = FirstSeenUtc<\/code> (set in constructor to DateTime.UtcNow)<\/li>\nm_currentPollCycleUtc<\/code> was set!<\/li>\n<\/ul>\n<\/li>\nMarkAsNeedsCompletionQuery<\/code><\/li>\n<\/ul>\n<\/li>\n<\/ol>\n\n
HandleOpenOrder<\/code>. Their LastSeenInPollUtc<\/code> is set to the time of creation.<\/li>\nGetOrAddOrderListTopic<\/code> is called, it:\n\n
BuildOrderListCsv()<\/code> (returns \u201c12345\u201d correctly)<\/li>\nEnsureOrdersRequested()<\/code> which sets m_currentPollCycleUtc = DateTime.UtcNow<\/code><\/li>\n<\/ul>\n<\/li>\nHandleOpenOrderEnd<\/code> is called:\n\n
m_currentPollCycleUtc > DateTime.MinValue<\/code> (true!)<\/li>\norderInfo.LastSeenInPollUtc < m_currentPollCycleUtc<\/code><\/li>\nLastSeenInPollUtc<\/code> (set during HandleOpenOrder<\/code>) may be BEFORE m_currentPollCycleUtc<\/code> (set during EnsureOrdersRequested<\/code>)<\/li>\nMarkAsNeedsCompletionQuery<\/code> is called<\/li>\nRemaining == 0<\/code> (not set in test), it sets NeedsCompletionQuery = true<\/code><\/li>\nRemaining > 0<\/code> (100 shares, 0 filled), so it just sets NeedsCompletionQuery = true<\/code> without changing status<\/li>\n<\/ul>\n<\/li>\n<\/ol>\nHandleOpenOrderEnd<\/code>, and the assertion is failing with an empty string!<\/p>\nPublishOrderListsIfAny()<\/code>:<\/p>\n\n
BuildOrderListCsv()<\/code> which returns \u201c12345\u201d<\/li>\nHandleOpenOrderEnd()<\/code> which calls PublishOrderListsIfAny()<\/code><\/li>\nPublishOrderListsIfAny()<\/code> calls BuildOrderListCsv()<\/code> again – this should still return \u201c12345\u201d<\/li>\n<\/ol>\nvar order1 = new Order { OrderId = 12345, PermId = 12345, Account = "TEST", Action = "BUY", TotalQuantity = 100 };\norderManager.HandleOpenOrder(12345, contract, order1, new OrderState { Status = "Submitted" });\n<\/code><\/pre>\nlistTopic.TopicValue?.ToString() ?? ""<\/code> AFTER calling HandleOpenOrderEnd()<\/code>.<\/p>\nvar entry = m_orderListTopics.GetOrAdd(key, _ =>\n{\n wasAdded = true;\n return new OrderListTopic(descriptor, topicId);\n});\n\ncreated = wasAdded;\n\nif (!wasAdded)\n{\n entry.Topic.AddTopicId(topicId);\n}\nelse\n{\n \/\/ Seed with current order list\n if (m_ordersSnapshotCompleted || !m_ordersRequested)\n {\n var csv = BuildOrderListCsv(descriptor);\n entry.Topic.SetTopicValue(csv);\n entry.LastValue = csv;\n }\n else\n {\n entry.Topic.SetTopicValue(string.Empty);\n entry.LastValue = string.Empty;\n }\n \n EnsureOrdersRequested();\n}\n<\/code><\/pre>\nm_ordersRequested<\/code> has been set to true<\/code> by some other code path before this call, then the condition !m_ordersRequested<\/code> is FALSE, and m_ordersSnapshotCompleted<\/code> is also FALSE, so we go to the else<\/code> branch and seed with empty string!<\/p>\nm_ordersRequested<\/code> be set before the test calls GetOrAddOrderListTopic<\/code>? Let me check if there\u2019s something in the test environment or if another test could affect this\u2026<\/p>\nm_ordersRequested = true<\/code> somehow\u2026 but wait, each test creates a NEW OrderManager instance, so that\u2019s not it.<\/p>\nShowFilledOrdersConfig<\/code>:<\/p>\nshowFilledOrdersOverride: false<\/code> so that\u2019s not the issue.<\/p>\nIsTerminalStatus<\/code>:<\/p>\n