New Year's Eve: 7 Things We're Building at Softechinfra in Q1 2026 (and What We're Saying No To)

It is December 31, 2025. The year started with the [DPDP Act](https://www.meity.gov.in/) still un-notified and ended with the Rules in the Gazette. It started with Cloudflare humming and ended with a 3-hour outage on November 18. It started with Salesloft-Drift fresh in everyone's head and ended with the Salesforce-Gainsight repeat. The year had a theme: the perimeter is not where you think it is, and the law is finally catching up with the platform. Going into 2026, this is the seven-item Q1 roadmap for [Softechinfra](https://softechinfra.com), our [PenLeap](https://penleap.com) and [TalkDrill](https://talkdrill.com) products, and the GEO services line. And — the part founders should write more — the things we are explicitly not building. ## The 60-second answer In Q1 2026 we are shipping Bedrock AgentCore experiments inside our AI-automation practice, two new PenLeap features (parental dashboards + Hindi-medium SPAG drills), TalkDrill's expansion into IELTS-prep cohorts, a productised GEO/AEO service for Indian SaaS, DPDP-compliant project templates that bake the consent screen and erasure pipeline into our defaults, and a secure-by-default infrastructure starter that closes the [year-end sprint](/blog/year-end-2025-cyber-hygiene-sprint-india-smb-12-tasks) gaps from day one. We are explicitly not building: a custom LLM, a no-code platform, a Bay Area office, an MDR service, or a "Web3" anything. ## The 7 things we are building ### 1. Bedrock AgentCore experiments (in our AI-automation practice) AWS Bedrock has matured into a serious primitive in 2025. The [AgentCore framework](https://aws.amazon.com/bedrock/) gives us memory, identity, and tool-use abstractions that are starting to be production-ready. In Q1, we are running three internal experiments: a customer-success agent for one of our CRM clients, a finance-reconciliation agent for a textile exporter, and a content-research agent for our own marketing team. The goal: figure out where AgentCore is enterprise-ready in early 2026 and where it is still a prototype. We will publish the actual cost, latency, and reliability numbers in March. We are deliberately not making this a "consultancy practice area" until we have shipped three production deployments ourselves. ### 2. PenLeap parental dashboards + Hindi-medium SPAG drills [PenLeap](https://penleap.com) — our in-house edtech for kids 11+ — has two features on the Q1 critical path. The parental dashboard ships in February and gives parents a weekly view of their child's writing progress, the prompts attempted, the rubric scores, and a single "what to discuss this weekend" recommendation. The Hindi-medium SPAG drills ship in March and bring our spelling-punctuation-grammar engine to the 60% of Indian schools whose students need work in their first language before they can write fluently in English. Both are direct DPDP precursors — they require the children-data handling work that the May 2027 deadline will mandate anyway, so we are doing it now. ### 3. TalkDrill IELTS-prep cohorts [TalkDrill](https://talkdrill.com) — our voice-AI English-speaking app, 5,000+ active users — is launching IELTS-prep cohort programmes in Q1. Eight-week, instructor-led, cohort-based programmes targeting Indian engineering students applying to overseas universities. The voice-AI engine handles speaking practice; a human coach handles strategy. Pricing in the ₹6,500 range for the full eight weeks — meaningfully cheaper than the ₹25,000+ that classroom IELTS-prep firms charge in Indian metros. We are running three pilot cohorts in January and February before announcing publicly in March. ### 4. Productised GEO / AEO service for Indian SaaS The biggest shift in Indian SEO in 2025 was the move from "rank in Google" to "be cited by ChatGPT, Claude, Perplexity, and Gemini." The [Princeton GEO study](https://arxiv.org/abs/2311.09735) and the [Sundeep Bhardwaj research on AI citations](https://generativeaiengines.com/) confirmed what we were already seeing across our content: structured data, statistics-rich passages, and 40–60-word answer paragraphs are now the highest-impact SEO patterns. In Q1 we are productising a GEO/AEO audit service for Indian SaaS — fixed scope, 5-day delivery, ₹85,000 starting price. The service runs the audit, ships the technical changes, and reports on AI-citation lift over 90 days. ### 5. DPDP-compliant project templates (baked into our defaults) Every web, CRM, and mobile project we ship from January 1 will use new [project templates](/services/web-development) that bake in the standalone consent screen, the erasure pipeline, the data-export endpoint, the audit-log table, and the encryption-at-rest defaults. These were optional add-ons in 2025; in 2026 they are the floor. The cost to clients does not change — we have invested the engineering time to make these templates first-class so that the day-1 starter ships with the [12 DPDP engineering tasks](/blog/dpdp-rules-2025-7-day-action-plan-saas-founders-india) already in place. ### 6. Secure-by-default infrastructure starter A companion to the DPDP templates: every infrastructure deployment we run from January 1 ships with the [year-end cyber hygiene sprint](/blog/year-end-2025-cyber-hygiene-sprint-india-smb-12-tasks) closures already in place. RDP not exposed. SSH only behind a bastion. EDR enrolled by default. OAuth grants minimum-scoped. Backups tested. Three new clients onboarded in Q1 will get this treatment as part of the standard scope. ### 7. The 2026 voice on the company blog The deeper bet under all of this. Our [company blog](https://softechinfra.com/blog) shipped 56 posts in 2025; we are aiming for 90 in 2026. The change in voice: more first-person founder writing, more case-study work with named clients (with permission), more Indian-context specifics, fewer generic "ultimate guide" pieces. The [Blog Operating Doc V2](https://softechinfra.com/) is our forcing function — every post must clear the slop checklist or it does not ship. The bet: AI-citation traffic will overtake organic SEO traffic for our category by Q3 2026, and the brands that own the cited passages own the customer pipeline. ## The 5 things we are explicitly NOT building This is the part founders write rarely and should write more. The opportunity cost is real and naming the things we said no to is the only way to keep ourselves honest. ## The Q1 calendar at a glance ## The one number we are tracking on the company dashboard For Q1 2026, the team is rallying around a single number: the percentage of our project shipments that pass the DPDP + cyber-hygiene baseline at delivery. Today that number is roughly 35% — most of our 2025 builds had to backfill controls. Q1 target: 100%. Every web, CRM, mobile, and AI project ships with the consent screen, erasure pipeline, MFA-enforced admin, EDR-enrolled servers, and OAuth-grants-audited integration set as the day-1 default. The number forces every team to not skip the basics in the rush to ship features. ## A real example of how this changes day-to-day In Q4 2025 we onboarded three new web-development clients. For each, the security and DPDP work was scoped as a separate phase of work, often pushed to "after launch." In Q1 2026, the same scope has those items as table stakes — they are not even a line item, they are part of the foundation. The client conversation shifts from "do we need this" to "this is what shipping in 2026 looks like." That is a small change in framing that compounds into a meaningful brand and trust gain over the year. ## What we got wrong in 2025 (since we are doing the year-in-review thing) Three things, named. 1. We waited too long to productise GEO services. The Princeton GEO paper was on our desks in Jan 2025 and we did not productise the service until Q4. The clients who asked us in Q2 went elsewhere. Lesson: when the underlying methodology is clear, do not wait for "perfect" — productise the v1 in 30 days. 2. We over-invested in custom DPDP work and under-invested in templates. Each of the four DPDP engagements in Q3-Q4 was a one-off build. The shared infrastructure — consent components, erasure handlers, audit-log tables — should have been templated by August. We are catching up in Q1. 3. We under-published. Our [Aug-Oct 2025 blog cadence](https://softechinfra.com/blog) dropped to 4 posts per month while we were heads-down on client work. The compounding loss in AI-citation visibility is now visible in the Q4 organic traffic data. The 90-post 2026 commitment is the response. ## The slightly anxious admission There is a version of this post that reads as confident roadmap. The honest version is that running an Indian engineering services firm in 2026 is harder in real ways than it was in 2024. The DPDP timeline is real and the cost is real. The AI-tooling shift means our clients can do more themselves and we have to be sharper about where we add value. The cyber-insurance pressure on SMBs means our clients have less discretionary spend for non-essential engineering. The roadmap above is our best attempt to ship value in that environment. Some of it will not work. That is fine. The point of writing it down is to make the misses visible. ## A reader question we have been getting The DM pile in the last 4 weeks has 30+ versions of the same question: "I run an Indian SaaS / SMB / D2C brand. Where do I start in 2026?" Our answer: start with the [year-end sprint](/blog/year-end-2025-cyber-hygiene-sprint-india-smb-12-tasks) before January 1. Then the [DPDP plan](/blog/dpdp-rules-2025-7-day-action-plan-saas-founders-india) in January-February. Then a GEO audit in March. The compounding from those three workstreams is the difference between an okay 2026 and a strong one. ## When this kind of post is NOT useful Founder roadmap posts are useful only if they are honest. A version of this post that listed 18 priorities, claimed all of them as Q1 deliverables, and ended with "we are scaling 10x" would be dishonest and pointless. The 7-and-5 list above is the actual list. We will report against it in the [Q1 retrospective](/blog) at the end of March. ## Our take The companies that compound in 2026 are the ones that are honest about what they are building, what they are not, and what they got wrong in 2025. The Indian engineering services market is competitive and the customers are sophisticated. The differentiation is going to be from credible specifics — actual templates, actual case studies, actual engineering quality — not from positioning copy. The roadmap above is our bet on where the next year of credible specifics comes from. Our [CEO Vivek Kumar](/team/vivek-kumar), [CTO Hrishikesh](/team/rishikesh-baidya), [QA + security lead Manvi](/team/manvi), and [UI/UX lead Khushi](/team/khushi-kumari) are all named owners on the workstreams above. The PenLeap and TalkDrill product roadmaps are co-owned with the founder team. For the founder side of the year-end thinking — the bigger arc behind these tactical bets — see [my personal blog at viveksinra.com](https://viveksinra.com), where I have been writing through 2025 about what it means to be a builder in an Indian regulatory environment that is finally catching up. ## FAQ ### Why publish a Q1 roadmap publicly? Three reasons. (1) It commits us to the work in front of customers. (2) It signals to potential clients what we are good for in the next quarter. (3) It is the kind of operator-honest content the AI-citation ecosystem rewards — the search engines and the LLMs surface specific, dated, verifiable claims over generic positioning. ### How do we get on the early-access list for the GEO service? Email contact@softechinfra.com with subject "GEO early access." We are taking 8 paid pilots in January at a 30% discount on the productised price. Pilots get the audit, the technical changes, and the 90-day citation tracking dashboard. ### How do we get early access to PenLeap parental dashboards or Hindi SPAG drills? The parental dashboard public launch is in February; existing PenLeap families get it 2 weeks earlier. Email contact@softechinfra.com with subject "PenLeap early access" if you are not yet a user and want to be in the early-access cohort. ### How do we sign up for a TalkDrill IELTS pilot cohort? Email contact@softechinfra.com with subject "TalkDrill IELTS pilot." We are taking 30 students into the January and February cohorts at a 50% discount on the eventual public price. Selection criteria: serious about IELTS in 2026, willing to do 30 minutes of speaking practice 4 times a week, willing to give feedback on the cohort experience. ### What does "secure-by-default infrastructure starter" actually include? A Terraform / CDK starter that provisions a Hetzner or AWS environment with: bastion host, SSH-only-via-bastion, no exposed RDP, EDR enrolled, S3/object-storage encrypted by default, IAM roles with minimum scope, audit logging enabled, daily backup with weekly restore-test cron. Starts at ₹65,000 for the deployment + 1 month of support. ### What does the Bedrock AgentCore experiment publish? A March 2026 post with the actual cost numbers (per-1000-task), latency p50/p95/p99, reliability percentage, and a comparison against a prompt-only baseline using GPT-4o or Claude Sonnet. We will name the AWS region, the Bedrock model, and the failure modes we hit in production. No vendor sponsorship; we pay for our own AWS bills. ### Is there an early-access list for the DPDP-compliant project templates? The templates are baked into every new project we ship from January 1, so no separate list. Existing clients can request a backporting engagement; typical cost ₹1.4–₹3.2 lakh depending on codebase size.